US20080152201A1 - Efficient Propagation for Face Annotation - Google Patents
Efficient Propagation for Face Annotation Download PDFInfo
- Publication number
- US20080152201A1 US20080152201A1 US12/047,247 US4724708A US2008152201A1 US 20080152201 A1 US20080152201 A1 US 20080152201A1 US 4724708 A US4724708 A US 4724708A US 2008152201 A1 US2008152201 A1 US 2008152201A1
- Authority
- US
- United States
- Prior art keywords
- face
- name
- images
- faces
- engine
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/54—Browsing; Visualisation therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/179—Human faces, e.g. facial parts, sketches or expressions metadata assisted face recognition
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99941—Database schema or data structure
- Y10S707/99944—Object-oriented database structure
- Y10S707/99945—Object-oriented database structure processing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
- Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y10S707/00—Data processing: database and file management or data structures
- Y10S707/99941—Database schema or data structure
- Y10S707/99948—Application of database or data structure, e.g. distributed, multimedia, or image
Definitions
- the subject matter relates generally to digital image management and more specifically to efficient propagation for face annotation.
- Face annotation is very important in the management of digital photo albums, because it provides information about who is in the photos and thus can help users organize and manage their digital albums. Although there are already many commercial products that try to assist with electronic photo album annotation, they all require human annotations, a tedious task very few users will perform.
- faces in photos are detected automatically by a face detector and if a user would like to annotate the faces, the conventional system calculates a candidate list of names from which the user annotates each face.
- the candidate list can be calculated according to a face's similarity with already annotated faces.
- the user might accept one of the recommendations, or instead enter a new name for the face.
- a candidate name list pops up to provide one or more recommendations.
- Systems and methods allow a user to select a group of images, such as digital photographs, and assign to the group the name of a person who is represented in each of the images. Then, the name is automatically propagated to the face of the person, each time the person's face occurs in an image.
- names and associations are shared between a browsing mode for simultaneously viewing multiple images and a viewer mode, for viewing images one at a time.
- the browsing mode can provide additional information to the viewer mode for generating a menu of candidate names for annotating a face in a single image.
- the viewer mode can provide manually annotated face information to the browser mode for facilitating automatic name propagation.
- identification of a person's face in multiple images can be accomplished not only by finding similarities in facial features but also by finding similarities in contextual features near the face in different images.
- the subject matter formulates exemplary name propagation as an optimization problem.
- An objective function is defined as the sum of similarities between each pair of faces of the same individual appearing in different photographs. Maximization of the objective function leads to the optimal solution for name propagation.
- name propagation accuracy can be improved if some faces have been previously annotated, and similarly, annotation accuracy can be improved if some photographs have already been associated with names.
- FIG. 1 is a diagram of an exemplary system for name propagation and face annotation.
- FIG. 2 is a block diagram of an exemplary face annotation engine.
- FIG. 3 is a diagram of exemplary mode integration for face annotation.
- FIG. 4 is a diagram of exemplary face annotation in a browsing mode.
- FIG. 5 is a diagram of exemplary face annotation in a viewer mode.
- FIG. 6 is a diagram of exemplary name propagation.
- FIG. 7 is a diagram of exemplary name dangling at an image level.
- FIG. 8 is a flow diagram of an exemplary method of face annotation.
- Described herein are systems, engines, user interfaces, and methods for annotating digital images in a collection, such as a family photo album.
- the subject matter can automatically propagate labels, such as names, from the level of a batch of images to the level of a visual item within the images, even though the item may vary somewhat between images.
- the subject matter provides an automatic or semi-automatic approach to annotate faces in a typical digital family photo album, which usually contains a limited number of approximately 10-100 persons who appear frequently.
- the systems and methods for annotating digital images in a collection can allow a user to select a batch of digital photos, apply to the batch the name of a person who appears in each photo, and then propagate the name to a face that is common to each photo in the selection.
- the systems and methods infer correspondence between name and face. In one implementation, this saves a human user from having to manually name people shown in photographs. Rather than open each photograph, a user can look at a batch of thumbnails to see who is in each photograph, and use a mouse to select multiple photographs with a particular individual that the user wants to label. Since browsing thumbnails has now become common in most photo management systems, users can annotate on a photograph level, instead of on a “face-by-face” level, i.e., users can efficiently label a group of photographs in a batch manner.
- FIG. 1 shows an example environment for implementing the systems and methods.
- a computing device 100 includes an exemplary face annotation engine 102 .
- the face annotation engine 102 may assist in managing one or more digital image collections 104 via user interfaces 106 and 106 ′, and can apply names to the faces that appear in the image collection 104 .
- a user selects a set of images from the image collection 104 (i.e., “multi-selects” images). In FIG. 1 , selected images are shown as having a darker image border on than non-selected images. The user then provides a name 108 for a person whose face is common to the selected images.
- the face annotation engine 102 matches the provided name 108 to the correct face in the selected images using a similarity measure function to compare facial and contextual features information, to be discussed more fully below.
- the face annotation engine 102 can then propagate the name to other images in which the face appears, for example, to other non-selected images in the image collection 104 or to new images added to the image collection 104 .
- the face annotation engine 102 manages face annotation across different modes of viewing the image collection 104 , exploiting advantages of each viewing mode for benefiting face annotation in the other mode(s). For example, given a browsing mode that displays thumbnail images and a “viewer” mode that displays single images one at a time, as in a slideshow, these two modes can accomplish face annotation synergistically.
- a drop-down menu can present the user with naming choices. Once a name is associated with a face in this manner, then this association can be used in the browsing mode to propagate the name to other images bearing the same face. Likewise, name propagation that occurs in the browsing mode can also facilitate generating the names to populate the drop-down menu in the viewer mode.
- FIG. 2 shows the exemplary face annotation engine 102 of FIG. 1 in greater detail.
- the face annotation engine 102 in FIG. 2 is meant to provide one example arrangement for the sake of overview. Many other arrangements of the illustrated components, or similar components, are possible. It is worth noting that the face annotation engine 102 of FIG. 2 can be executed in hardware, software, or combinations of hardware, software, firmware, etc.
- the exemplary face annotation engine 102 includes a user interface 202 , an image multi-select engine 204 , an image collection manager 206 , a name propagation engine 208 , and a mode integrator 210 .
- the user interface 202 includes a display 212 and a name inputter 214 .
- the image multi-select engine 204 may include a database or storage area of selected images 216 .
- the image collection manager 206 stored the images 218 and a labeled faces list 220 .
- the name propagation engine 208 includes an images reader 222 , a face finder 224 , and a face identifier 226 .
- the face identifier 226 further includes an images comparator 228 and a (name-to-face) correspondence engine 230 .
- the correspondence engine 230 includes a similarity measure engine 232 and a contextual features engine 234 .
- the mode integrator 210 further includes a browsing mode manager 236 , and a viewer mode manager 238 that may include a menu generator 240 .
- the face annotation engine 102 integrates name annotation activities and shared data between the browsing mode manager 236 and the viewer mode manager 238 . That is, as shown in FIG. 3 , the mode integrator 210 integrates face annotation that occurs in a browsing mode 302 (in which multiple of the images 218 are presented as thumbnails in the display 212 ) with face annotation that occurs in a viewer mode 304 (in which the images 218 are presented on the display 212 one at a time). The two modes do not necessarily occur at the same time.
- the viewer mode manager 238 allows a user to move a mouse, for example, onto a face 306 in a displayed image 308 , and the menu generator 240 creates a candidate name list 310 that can appear to provide a labeling recommendation.
- the user can either select a name from the name list 310 , or enter a new name via the name inputter 214 for the selected face 306 .
- the menu generator 240 generates the candidate name list 310 from historical labeling and/or name propagation results, i.e., the labeled faces list 220 .
- the multi-select engine 204 allows a user to select a number of the images 218 , e.g., thumbnails, that each include a particular person, and then enter the person's name 108 or other label via the name inputter 214 , in order to globally perform comprehensive (e.g., one-click) assignment of the name 108 to the face of the person in each of the multiple selected images 216 .
- the name propagation engine 208 tries to infer the correspondence between the input name 108 and the face as it exists in its various forms in the different selected images 216 , i.e., tries to propagate the name 108 from image level to face level. Exemplary name propagation will be discussed in more detail further below.
- the browsing mode manager 236 may provide even more effectiveness by offering similar face retrieval. Similar face retrieval allows users to search for similar faces by specifying either a face or a name. Then multiple faces can be annotated in a batch manner. In either manner of browsing, batch annotation can be conducted.
- the mode integrator 210 allows the results from one mode of viewing and annotating to benefit the viewing and/or annotating provided by the other mode. That is, the labeled faces list 220 can be used and developed by either mode. The labeled faces list 220 benefits from both the name propagation of the browsing mode 302 and the more manual user-labeling of faces of the viewer mode 304 .
- some implementations of the face annotation engine 102 may also include components to perform a feature known as “smart thumbnailing” (See, Wang M. Y., Xie X., Ma W. Y., Zhang H. J., “MobiPicture—Browsing Pictures on Mobile Devices”, in ACM International Conference on Multimedia, Berkeley, 2003).
- thumbnail images are generated by directly downsampling original, larger images 218 .
- thumbnails generated by this conventional approach are difficult to recognize, especially when the thumbnails are very small.
- the browsing mode manager 236 or other components may automatically generate the smart thumbnails.
- the browsing mode manager 236 when a user moves the mouse onto a thumbnail, the browsing mode manager 236 automatically switches to display the most informative part of the thumbnail image, i.e., face areas, to the user. In one implementation, if the mouse pointer hovers on the thumbnail image for few seconds, the browsing mode manager 236 automatically animates the image with an optimal browsing path to let the user browse through the different parts of the image.
- one implementation of the name propagation engine 208 as shown in FIG. 2 tries to propagate a name 108 (that has been applied to a group of selected images 216 ) to a face that appears in the selected images 216 .
- the images reader 222 accesses digital data of each of the selected images 216 while a face finder 224 seeks to detect each face that occurs in each image, including faces to which the name 108 will not be assigned.
- the face identifier 226 then tries to find that face in each of the selected images 216 to which the name 108 does apply.
- the face identifier 226 extracts facial appearance features based on face detection results. Because face appearance features are most reliable when extracted from frontal faces, a texture-constrained active shape model is applied to determine if an input face is in frontal view. (See, Yan S. C., et al., “Texture-Constrained Active Shape Models”, Proceedings of The First International Workshop on Generative - Model - Based Vision , Copenhagen, Denmark, May, 2002).
- the face identifier 226 treats the face appearance feature as a missing feature. After this process, the face identifier 226 aligns each face into standard normalized form to remove the variations in translation, scale, in-plane rotation, and slight out-of-plane rotation.
- facial appearance features are extracted from normalized gray face images.
- the images comparator 228 or other control fabric of the exemplary face identifier 226 can include logic to ignore one or more of the images in the event of mistaken selection. For example, if all of the selected images 216 except one appear to include a face that should be associated with the name 108 , then the one may be discarded as mistakenly selected, or the face unidentifiable.
- the images comparator 228 may also determine that a qualifying face occurs in each of the selected images 216 .
- the name-to-face correspondence engine (“correspondence engine”) 230 employs various techniques, such as a similarity measure function performed by the similarity measure engine 232 , to propagate a name 108 from image level to face level.
- the similarity measure engine 232 may use facial features to identify similar faces, but contextual features may also be used in the similarity measure as now described.
- the correspondence engine 230 uses contextual image features in addition to facial appearance features to represent each face. That is, similarities are not only drawn between facial features of candidate faces, but also between the visual contexts near each face in an image. This is especially useful when the images are digital photos, because after multiple photos of a limited group of people are taken within a short time span—those being photographed are typically wearing the same clothes across multiple photos. These photos are often downloaded from a digital camera and stored in the same folder. A browsing mode manager 236 displays these photos simultaneously, and if selected, correspondence engine 230 more easily finds a face with higher confidence in each photo due to the same clothes appearing under the face in each photo.
- the contextual features engine 234 extracts contextual features from an extended face region, such as the example extended face region 502 .
- the contextual features engine 234 can use, for example, a 44-dimensional banded auto-correlogram and/or a 14 dimensional color texture moment as described, respectively, by Huang J., Kumar S. R., Mitra M., Zhu W. J. and Zabih R., in “Image indexing using color correlograms”, IEEE Conference on Computer Vision and Pattern Recognition, p. 762, 1997; and Yu H., Li M., Zhang H., and Feng J., “Color texture moment for content-based image retrieval”, Proc. IEEE Intl Conf.
- the contextual features engine 234 divides the extended face region 502 into, for example, 2 ⁇ 1 blocks, and extracts local regional features to capture the structural information of body patches.
- the contextual features engine 234 can restrict the difference in date attributes of two images to be within a certain time period, such as two days, when comparing images for their contextual similarity.
- Similarity such as facial similarity and/or similarity of contextual features, is defined as maximum a posteriori (MAP) estimation in Zhang L., Chen L., Li M., Zhang H. “Automated annotation of human faces in family albums”, ACM International Conference on Multimedia, pp. 335-358, Berkeley, 2003; and Lei Zhang, Yuxiao Hu, Mingjing Li, Weiying Ma, Hongjiang Zhang, “Efficient Propagation for Face Annotation in Family Albums,” MM'04, Oct. 10-16, 2004, ACM Multimedia, New York, 2004. Both of these resources are incorporated by reference herein.
- the correspondence engine 230 and specifically the similarity measure engine 232 , perform name propagation by solving a MAP similarity optimization problem.
- p( ⁇ I ) and p( ⁇ E ) are the a priors
- ⁇ i ( ⁇ i1 ⁇ i2 )
- ⁇ E ) are the likelihoods for a given difference ⁇ i .
- This similarity function integrates multiple features into a Bayesian framework. In case there are missing features, marginal probability is used so that samples that have missing features can be compared with those having the full feature set to ensure a non-biased decision. Based on this similarity measure, candidate names for a given unknown face can be derived by statistical learning approaches, such as K nearest neighborhood algorithm.
- F i ⁇ F 1 i , . . . , F Ci i ⁇ be the faces (individuals) in I i , the selected images 216 .
- the similarity measure engine 232 defines an objective function as the sum of similarities between each pair of faces of the same individual in different photographs, as it is assumed that these selected images 216 share the same property in terms of the specified individual associated with the name 108 , and the faces of the one individual should be similar with each other. Therefore, the objective function can be formulated as:
- the similarity measure engine 232 maximizes the objective function to obtain the optimal solution for name propagation:
- ⁇ f 1 , f 2 , ... ⁇ ⁇ f N ⁇ arg ⁇ ⁇ max f 1 , f 2 , ... ⁇ ⁇ f N ⁇ ⁇ Sim ⁇ ( f 1 , f 2 , ... ⁇ ⁇ f N ) ( 3 )
- FIG. 6 shows an example optimal solution for name propagation.
- Four selected images 216 (four photographs) are assigned “Peter” as a name 108 . That is, via the name inputter 214 , a user informs the face annotation engine 102 that each one of the four photographs contains “Peter”.
- the correspondence engine 230 associates these faces with the “Peter” name 108 and adds the associations to the labeled faces list 220 .
- the similarity measure engine 232 uses an iterative optimization technique. It is an NP-hard problem (a nondeterministic polynomial time problem) to find the globally optimal solution as defined in Equation (3) above for name propagation. (A problem is NP-hard if solving it in polynomial time—the set or property of problems which can be solved by a known polynomial-time algorithm—would make it possible to solve all problems in class NP in polynomial time.) For comparison, a brute force approach would try to enumerate all possible solutions and select the optimal solution in terms of maximal similarity defined in Equation (3) above. However, this enumerative approach would result in a combination explosion problem.
- NP-hard problem a nondeterministic polynomial time problem
- the similarity measure engine 232 can use an iterative approach to the optimization problem.
- the correspondence engine 230 first selects an initial solution, and then iteratively and greedily adjusts faces in each of the selected images 216 to obtain a solution with larger similarity, until the increasing similarity reaches or approaches stability.
- the similarity measure engine 232 performs the technique, following these steps:
- the similarity measure engine 232 calculates a new similarity sum Sim as:
- the similarity measure engine 232 exits the steps of the technique and outputs solution ⁇ 1 , ⁇ 2 , . . . , ⁇ N ⁇ .
- the similarity measure engine 232 greedily searches for new solution and guarantees that objective similarity sum is monotonously increased.
- the similarity measure engine 232 can at least reach local optima (or maxima).
- the similarity measure engine 232 greatly reduces the computational complexity from O(C 1 *C 2 * . . . *C N ) to O(m(C 1 +C 2 + . . . +C N )), where m is the iteration number.
- the correspondence engine 230 can flexibly select an enumerative approach or an iterative approach in accordance with the estimation of computational complexity.
- the system calculates a confidence score for each face as follows:
- the confidence score actually reflects the uniqueness of face ⁇ i in the photo. If the condition Conf i ⁇ T reject , a predetermined threshold for rejection, is true, then the name 108 will not be propagated to this face, and instead, the name is kept only loosely associated with an image instead of a face, that is the name is “dangled” with the image. Dangling names are still useful for both photograph searching and further face annotation. For a photographic search, because the names have been associated with photographs, it is straightforward to find these photographs by the associated names. For face annotation in viewer mode 304 , if there is a name associated with a photograph that has a face that the user wants to label, the candidate name list generation will be more accurate, as the dangling name is actually a strong prior.
- the name propagation engine 208 in browsing mode 302 there may be some additional inputs from viewer mode 304 , i.e. labeled faces in some images. For example, among the photographs selected by a user, there may be some previously labeled faces, e.g., in the labeled faces list 220 . In this case, the iterative optimization performed by the similarity measure engine 232 is adapted to utilize such additional information.
- Equation (4) is modified to include the influence from P ⁇ .
- Equation (5) is also modified to include the influence from P ⁇ .
- the view mode manager 236 can adopt an ad hoc strategy by adjusting the prior of dangling name P in photographs I i as follows:
- the prior of “Peter” is estimated to amount to only 10% from previous labeling results and one name “Peter” is dangling on photograph (a) 702 in FIG. 7 .
- the prior that face A is “Peter” is 33%. So, 33% is taken as the prior probability of “Peter” in generating the candidate name list for unknown face A.
- multiple dangling names are also useful for annotation, because any face being annotated reduces uncertainties in the corresponding photograph.
- photograph (b) 704 in FIG. 7 there are two dangling names associated with photograph (b) 704 . Once face A is annotated as “Peter”, the name to be associated with face B is immediately determined to be “Tom”.
- FIG. 8 depicts an exemplary method 800 to automatically annotate faces in an image collection.
- the operations are summarized in individual blocks. The operations may be performed by hardware, software, or combinations of both, for example, by components of an exemplary face annotation engine 102 .
- digital images are selected.
- a browsing mode implemented on a computing device may display thumbnail images of photos stored in a folder on the hard drive. From the entire array of displayed images, a user can select images at will.
- the user selects images that have one or more people in each selected image. The selection itself chooses a set of images that each include a person whose face will be annotated across the selected photos.
- the set of images is then named with the name or other moniker of the person common to all the selected images.
- the exemplary method does not know whose face in the selected photographs should be associated with the input name.
- the name is propagated to a face of the individual represented in each of the images.
- the exemplary method performs a similarity measure function using the features of pairs of faces in the different pictures.
- the similarity measure can be greatly enhanced by adding contextual features to the faces, e.g., a person's clothes.
- Computer subdirectory folders of downloaded photos from a digital camera are often limited to a group of photos that were all taken within a short time period, for example at one event. Thus, the human subjects of the photos often wear the same clothes. This extra-facial information enhances similarity comparisons between face pairs.
- the exemplary method includes defining an objective function as a sum of similarities between each pair of faces of the individual in different selected images, and maximizing the objective function in order to associate the name with the face in each image. Once an optimal solution is calculated, the name is applied to the face as it is represented in each selected photo.
- the annotated faces and names are stored to facilitate future name propagation and face annotation.
- program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types.
- program modules may also be practiced in distributed communications environments where tasks are performed over wireless communication by remote processing devices that are linked through a communications network.
- program modules may be located in both local and remote communications device storage media including memory storage devices.
Abstract
Systems, engines, user interfaces, and methods allow a user to select a group of images, such as digital photographs, and assign to the group of images the name of a person who is represented in each of the images. The name is automatically propagated to the face of the person, each time the person's face occurs in an image. In one implementation, names and associations are shared between a browsing mode for viewing multiple images at once and a viewer mode, for viewing one image at a time. The browsing mode can provide a menu of candidate names for annotating a face in a single image of the viewer mode. Likewise, the viewer mode can provide annotated face information to the browser mode for facilitating name propagation. Identification of a person's face in multiple images can be accomplished not only by finding similarities in facial features but also by finding similarities in contextual features near the face in different images.
Description
- This continuation application claims priority to U.S. patent application Ser. No. 11/111,696 to Zhang et al., entitled, “Efficient Propagation for Face Annotation,” filed Apr. 21, 2005 and incorporated herein by reference.
- The subject matter relates generally to digital image management and more specifically to efficient propagation for face annotation.
- Face annotation is very important in the management of digital photo albums, because it provides information about who is in the photos and thus can help users organize and manage their digital albums. Although there are already many commercial products that try to assist with electronic photo album annotation, they all require human annotations, a tedious task very few users will perform.
- Most conventional products offer a degree of image management for digital photo albums, but the image management is similar to the file management, which is based on the name, size, date/time and other properties of the image files. Face annotation is not provided. Only manual annotation of images is provided in conventional products.
- In one conventional product, faces in photos are detected automatically by a face detector and if a user would like to annotate the faces, the conventional system calculates a candidate list of names from which the user annotates each face. The candidate list can be calculated according to a face's similarity with already annotated faces. The user might accept one of the recommendations, or instead enter a new name for the face. In a typical scenario, if a user desires to label a face, the user moves a mouse onto the face, and a candidate name list pops up to provide one or more recommendations.
- Systems and methods allow a user to select a group of images, such as digital photographs, and assign to the group the name of a person who is represented in each of the images. Then, the name is automatically propagated to the face of the person, each time the person's face occurs in an image. In one implementation, names and associations are shared between a browsing mode for simultaneously viewing multiple images and a viewer mode, for viewing images one at a time. The browsing mode can provide additional information to the viewer mode for generating a menu of candidate names for annotating a face in a single image. Likewise, the viewer mode can provide manually annotated face information to the browser mode for facilitating automatic name propagation. In one implementation, identification of a person's face in multiple images can be accomplished not only by finding similarities in facial features but also by finding similarities in contextual features near the face in different images.
- The subject matter formulates exemplary name propagation as an optimization problem. An objective function is defined as the sum of similarities between each pair of faces of the same individual appearing in different photographs. Maximization of the objective function leads to the optimal solution for name propagation. To make the system more effective in annotation and propagation, name propagation accuracy can be improved if some faces have been previously annotated, and similarly, annotation accuracy can be improved if some photographs have already been associated with names.
-
FIG. 1 is a diagram of an exemplary system for name propagation and face annotation. -
FIG. 2 is a block diagram of an exemplary face annotation engine. -
FIG. 3 is a diagram of exemplary mode integration for face annotation. -
FIG. 4 is a diagram of exemplary face annotation in a browsing mode. -
FIG. 5 is a diagram of exemplary face annotation in a viewer mode. -
FIG. 6 is a diagram of exemplary name propagation. -
FIG. 7 is a diagram of exemplary name dangling at an image level. -
FIG. 8 is a flow diagram of an exemplary method of face annotation. - Described herein are systems, engines, user interfaces, and methods for annotating digital images in a collection, such as a family photo album. The subject matter can automatically propagate labels, such as names, from the level of a batch of images to the level of a visual item within the images, even though the item may vary somewhat between images. In one implementation, the subject matter provides an automatic or semi-automatic approach to annotate faces in a typical digital family photo album, which usually contains a limited number of approximately 10-100 persons who appear frequently.
- For example, the systems and methods for annotating digital images in a collection can allow a user to select a batch of digital photos, apply to the batch the name of a person who appears in each photo, and then propagate the name to a face that is common to each photo in the selection. Thus, the systems and methods infer correspondence between name and face. In one implementation, this saves a human user from having to manually name people shown in photographs. Rather than open each photograph, a user can look at a batch of thumbnails to see who is in each photograph, and use a mouse to select multiple photographs with a particular individual that the user wants to label. Since browsing thumbnails has now become common in most photo management systems, users can annotate on a photograph level, instead of on a “face-by-face” level, i.e., users can efficiently label a group of photographs in a batch manner.
-
FIG. 1 shows an example environment for implementing the systems and methods. In one implementation, acomputing device 100 includes an exemplaryface annotation engine 102. Theface annotation engine 102 may assist in managing one or moredigital image collections 104 viauser interfaces image collection 104. In one implementation of the subject matter, a user selects a set of images from the image collection 104 (i.e., “multi-selects” images). InFIG. 1 , selected images are shown as having a darker image border on than non-selected images. The user then provides aname 108 for a person whose face is common to the selected images. Theface annotation engine 102 matches the providedname 108 to the correct face in the selected images using a similarity measure function to compare facial and contextual features information, to be discussed more fully below. Theface annotation engine 102 can then propagate the name to other images in which the face appears, for example, to other non-selected images in theimage collection 104 or to new images added to theimage collection 104. - In another aspect of the subject matter, the
face annotation engine 102 manages face annotation across different modes of viewing theimage collection 104, exploiting advantages of each viewing mode for benefiting face annotation in the other mode(s). For example, given a browsing mode that displays thumbnail images and a “viewer” mode that displays single images one at a time, as in a slideshow, these two modes can accomplish face annotation synergistically. In the viewer mode, for example, a drop-down menu can present the user with naming choices. Once a name is associated with a face in this manner, then this association can be used in the browsing mode to propagate the name to other images bearing the same face. Likewise, name propagation that occurs in the browsing mode can also facilitate generating the names to populate the drop-down menu in the viewer mode. -
FIG. 2 shows the exemplaryface annotation engine 102 ofFIG. 1 in greater detail. Theface annotation engine 102 inFIG. 2 is meant to provide one example arrangement for the sake of overview. Many other arrangements of the illustrated components, or similar components, are possible. It is worth noting that theface annotation engine 102 ofFIG. 2 can be executed in hardware, software, or combinations of hardware, software, firmware, etc. - A list of exemplary engine components is now provided. The exemplary
face annotation engine 102 includes a user interface 202, an imagemulti-select engine 204, animage collection manager 206, aname propagation engine 208, and amode integrator 210. The user interface 202 includes adisplay 212 and aname inputter 214. Theimage multi-select engine 204 may include a database or storage area of selectedimages 216. Theimage collection manager 206 stored theimages 218 and a labeled faceslist 220. Thename propagation engine 208 includes animages reader 222, aface finder 224, and aface identifier 226. Theface identifier 226 further includes animages comparator 228 and a (name-to-face)correspondence engine 230. Thecorrespondence engine 230 includes asimilarity measure engine 232 and acontextual features engine 234. Themode integrator 210 further includes abrowsing mode manager 236, and aviewer mode manager 238 that may include amenu generator 240. - In one implementation, the
face annotation engine 102 integrates name annotation activities and shared data between thebrowsing mode manager 236 and theviewer mode manager 238. That is, as shown inFIG. 3 , themode integrator 210 integrates face annotation that occurs in a browsing mode 302 (in which multiple of theimages 218 are presented as thumbnails in the display 212) with face annotation that occurs in a viewer mode 304 (in which theimages 218 are presented on thedisplay 212 one at a time). The two modes do not necessarily occur at the same time. - In the viewer mode 304, the
viewer mode manager 238 allows a user to move a mouse, for example, onto aface 306 in a displayedimage 308, and themenu generator 240 creates acandidate name list 310 that can appear to provide a labeling recommendation. The user can either select a name from thename list 310, or enter a new name via the name inputter 214 for the selectedface 306. Themenu generator 240 generates thecandidate name list 310 from historical labeling and/or name propagation results, i.e., the labeled faceslist 220. - In the browsing mode 302, as shown in
FIG. 4 , themulti-select engine 204 allows a user to select a number of theimages 218, e.g., thumbnails, that each include a particular person, and then enter the person'sname 108 or other label via thename inputter 214, in order to globally perform comprehensive (e.g., one-click) assignment of thename 108 to the face of the person in each of the multiple selectedimages 216. In this browsing mode 302, thename propagation engine 208 tries to infer the correspondence between theinput name 108 and the face as it exists in its various forms in the different selectedimages 216, i.e., tries to propagate thename 108 from image level to face level. Exemplary name propagation will be discussed in more detail further below. - In either mode, generation of the labeled faces
list 220 greatly reduces users' efforts in face annotation and empowers users in indexing and searching their photo albums. Rather than opening each photograph in the viewer mode 304, it is often more convenient for users to view a collection of thumbnails to see who is in each photograph and perform the sweeping (e.g., one-click) assignment of a name to multiple photographs, as in the browsing mode 302. In most cases, users browse their albums in the order of the folder, which is closely related to time and event. Alternatively, thebrowsing mode manager 236 may provide even more effectiveness by offering similar face retrieval. Similar face retrieval allows users to search for similar faces by specifying either a face or a name. Then multiple faces can be annotated in a batch manner. In either manner of browsing, batch annotation can be conducted. - Since users frequently alternate between viewer mode 304, such as a sequential slideshow, and browsing mode 302 of a collection of thumbnails, it makes sense for the annotation that occurs in each mode to assist the other mode. The
mode integrator 210 allows the results from one mode of viewing and annotating to benefit the viewing and/or annotating provided by the other mode. That is, the labeled faceslist 220 can be used and developed by either mode. The labeled faceslist 220 benefits from both the name propagation of the browsing mode 302 and the more manual user-labeling of faces of the viewer mode 304. - It should be noted that some implementations of the
face annotation engine 102 may also include components to perform a feature known as “smart thumbnailing” (See, Wang M. Y., Xie X., Ma W. Y., Zhang H. J., “MobiPicture—Browsing Pictures on Mobile Devices”, in ACM International Conference on Multimedia, Berkeley, 2003). Conventionally, thumbnail images are generated by directly downsampling original,larger images 218. Often, however, thumbnails generated by this conventional approach are difficult to recognize, especially when the thumbnails are very small. To overcome this limitation, thebrowsing mode manager 236 or other components may automatically generate the smart thumbnails. In one implementation, when a user moves the mouse onto a thumbnail, thebrowsing mode manager 236 automatically switches to display the most informative part of the thumbnail image, i.e., face areas, to the user. In one implementation, if the mouse pointer hovers on the thumbnail image for few seconds, thebrowsing mode manager 236 automatically animates the image with an optimal browsing path to let the user browse through the different parts of the image. - As mentioned, one implementation of the
name propagation engine 208 as shown inFIG. 2 tries to propagate a name 108 (that has been applied to a group of selected images 216) to a face that appears in the selectedimages 216. To accomplish the propagation, theimages reader 222 accesses digital data of each of the selectedimages 216 while aface finder 224 seeks to detect each face that occurs in each image, including faces to which thename 108 will not be assigned. Theface identifier 226 then tries to find that face in each of the selectedimages 216 to which thename 108 does apply. - In one implementation, following an approach similar to that of Zhang L., Chen L., Li M., Zhang H. “Automated annotation of human faces in family albums”, in ACM International Conference on Multimedia, pp. 335-358, Berkeley, 2003 (incorporated herein by reference), the
face identifier 226 extracts facial appearance features based on face detection results. Because face appearance features are most reliable when extracted from frontal faces, a texture-constrained active shape model is applied to determine if an input face is in frontal view. (See, Yan S. C., et al., “Texture-Constrained Active Shape Models”, Proceedings of The First International Workshop on Generative-Model-Based Vision, Copenhagen, Denmark, May, 2002). If the face is not in frontal view, theface identifier 226 treats the face appearance feature as a missing feature. After this process, theface identifier 226 aligns each face into standard normalized form to remove the variations in translation, scale, in-plane rotation, and slight out-of-plane rotation. In one implementation, facial appearance features are extracted from normalized gray face images. - The images comparator 228 or other control fabric of the
exemplary face identifier 226 can include logic to ignore one or more of the images in the event of mistaken selection. For example, if all of the selectedimages 216 except one appear to include a face that should be associated with thename 108, then the one may be discarded as mistakenly selected, or the face unidentifiable. The images comparator 228 may also determine that a qualifying face occurs in each of the selectedimages 216. - The name-to-face correspondence engine (“correspondence engine”) 230 employs various techniques, such as a similarity measure function performed by the
similarity measure engine 232, to propagate aname 108 from image level to face level. Thesimilarity measure engine 232 may use facial features to identify similar faces, but contextual features may also be used in the similarity measure as now described. - In one implementation, for example, the
correspondence engine 230 uses contextual image features in addition to facial appearance features to represent each face. That is, similarities are not only drawn between facial features of candidate faces, but also between the visual contexts near each face in an image. This is especially useful when the images are digital photos, because after multiple photos of a limited group of people are taken within a short time span—those being photographed are typically wearing the same clothes across multiple photos. These photos are often downloaded from a digital camera and stored in the same folder. Abrowsing mode manager 236 displays these photos simultaneously, and if selected,correspondence engine 230 more easily finds a face with higher confidence in each photo due to the same clothes appearing under the face in each photo. - As shown in
FIG. 5 , thecontextual features engine 234 extracts contextual features from an extended face region, such as the example extendedface region 502. Thecontextual features engine 234 can use, for example, a 44-dimensional banded auto-correlogram and/or a 14 dimensional color texture moment as described, respectively, by Huang J., Kumar S. R., Mitra M., Zhu W. J. and Zabih R., in “Image indexing using color correlograms”, IEEE Conference on Computer Vision and Pattern Recognition, p. 762, 1997; and Yu H., Li M., Zhang H., and Feng J., “Color texture moment for content-based image retrieval”, Proc. IEEE Intl Conf. on Image Processing, September, 2002; which are incorporate herein by reference. The color moment feature in LUV color space (with luminance and chrominance components) can compensate if there is a lack of global color feature. In this exemplary implementation, thecontextual features engine 234 divides theextended face region 502 into, for example, 2×1 blocks, and extracts local regional features to capture the structural information of body patches. Thecontextual features engine 234 can restrict the difference in date attributes of two images to be within a certain time period, such as two days, when comparing images for their contextual similarity. - Similarity, such as facial similarity and/or similarity of contextual features, is defined as maximum a posteriori (MAP) estimation in Zhang L., Chen L., Li M., Zhang H. “Automated annotation of human faces in family albums”, ACM International Conference on Multimedia, pp. 335-358, Berkeley, 2003; and Lei Zhang, Yuxiao Hu, Mingjing Li, Weiying Ma, Hongjiang Zhang, “Efficient Propagation for Face Annotation in Family Albums,” MM'04, Oct. 10-16, 2004, ACM Multimedia, New York, 2004. Both of these resources are incorporated by reference herein. In one implementation, the
correspondence engine 230, and specifically thesimilarity measure engine 232, perform name propagation by solving a MAP similarity optimization problem. - For example, let F={ƒi|i=1, . . . , Nθ} denote the set of features to undergo similarity analysis, where each feature ƒi is a vector corresponding to a specific feature. By introducing two classes of face variations, intra-personal variations ΩI and inter-personal variations ΩE, and imposing an independent assumption for each feature ƒi, the similarity between two faces is defined as in
-
- where p(ΩI) and p(ΩE) are the a priors, Δƒi=(ƒi1−ƒi2), and p(Δƒi|ΩI) and p(Δƒi|ΩE) are the likelihoods for a given difference Δƒi. This similarity function integrates multiple features into a Bayesian framework. In case there are missing features, marginal probability is used so that samples that have missing features can be compared with those having the full feature set to ensure a non-biased decision. Based on this similarity measure, candidate names for a given unknown face can be derived by statistical learning approaches, such as K nearest neighborhood algorithm.
- Given the face similarity measure as defined in Equation (1) above, the
similarity measure engine 232 attempts to solve name propagation as an optimization problem, as follows. Assuming N selected images 216 (e.g., photos) at themulti-select engine 204, denoted by I={I1, I2, . . . , IN} with “,” (e.g. “Peter”) as thename 108 of these N selectedimages 216, each photo Ii contains , thename 108. Let Fi={F1 i, . . . , FCi i} be the faces (individuals) in Ii, the selectedimages 216. The name propagation problem is to select N faces denoted by Θ={ƒ1, ƒ2, . . . , ƒN} from N photos (one face from one photo), and assign thename 108 to these N faces. - In one implementation, the
similarity measure engine 232 defines an objective function as the sum of similarities between each pair of faces of the same individual in different photographs, as it is assumed that these selectedimages 216 share the same property in terms of the specified individual associated with thename 108, and the faces of the one individual should be similar with each other. Therefore, the objective function can be formulated as: -
- The
similarity measure engine 232 maximizes the objective function to obtain the optimal solution for name propagation: -
-
FIG. 6 shows an example optimal solution for name propagation. Four selected images 216 (four photographs) are assigned “Peter” as aname 108. That is, via thename inputter 214, a user informs theface annotation engine 102 that each one of the four photographs contains “Peter”. Based on the similarity measure between two faces described above, thesimilarity measure engine 232 tries to infer the correspondence between a face in each photo and the “Peter”name 108. There are 3×2×2×5=60 possible solutions in total for face selection, given the number of faces in each photo. But because faces “A3,” “B2,” “C2,” and “D1” are similar with each other (illustrated by similar shade), thecorrespondence engine 230 associates these faces with the “Peter”name 108 and adds the associations to the labeled faceslist 220. - In one implementation, the
similarity measure engine 232 uses an iterative optimization technique. It is an NP-hard problem (a nondeterministic polynomial time problem) to find the globally optimal solution as defined in Equation (3) above for name propagation. (A problem is NP-hard if solving it in polynomial time—the set or property of problems which can be solved by a known polynomial-time algorithm—would make it possible to solve all problems in class NP in polynomial time.) For comparison, a brute force approach would try to enumerate all possible solutions and select the optimal solution in terms of maximal similarity defined in Equation (3) above. However, this enumerative approach would result in a combination explosion problem. For example, if themulti-select engine 204 chose a set of twenty selectedimages 216 and there were two faces in each image, the number of possible solutions would be 220=1,048,576. To avoid combination explosion, thesimilarity measure engine 232 can use an iterative approach to the optimization problem. - The
correspondence engine 230 first selects an initial solution, and then iteratively and greedily adjusts faces in each of the selectedimages 216 to obtain a solution with larger similarity, until the increasing similarity reaches or approaches stability. In one implementation, thesimilarity measure engine 232 performs the technique, following these steps: - 1) The
similarity measure engine 232 accepts as givens, N selected images 216 {I1, I2, . . . , IN}, where each image Ii contains face set Fi={F1 i, . . . , FCi i} and similarity measure S(F1, F2). - 2) The
similarity measure engine 232 selects N faces from N images (one face from one image) as an initial solution Θ={ƒ1, ƒ2, . . . , ƒN}, and sets initial similarity sum, e.g., “Sim_Old”=0. - 3) For t=1, . . . , N, in image It, that is, in face set Ft, the
similarity measure engine 232 selects ƒt as: -
- 4) The
similarity measure engine 232 calculates a new similarity sum Sim as: -
- 5) If Sim—Sim_Old>ε, where ε is the pre-determined convergence threshold, then the
similarity measure engine 232 sets Sim_Old=Sim and goes back tostep 3, above. - 6) Otherwise, the
similarity measure engine 232 exits the steps of the technique and outputs solution {ƒ1, ƒ2, . . . , ƒN}. - In each iteration of the above steps, the
similarity measure engine 232 greedily searches for new solution and guarantees that objective similarity sum is monotonously increased. Thus thesimilarity measure engine 232 can at least reach local optima (or maxima). - The
similarity measure engine 232 greatly reduces the computational complexity from O(C1*C2* . . . *CN) to O(m(C1+C2+ . . . +CN)), where m is the iteration number. In practice, thecorrespondence engine 230 can flexibly select an enumerative approach or an iterative approach in accordance with the estimation of computational complexity. - Users usually expect correct and successful name propagation results. However, the
face annotation engine 102 may occasionally create an incorrect propagation result. As it is generally impractical to let users double-check propagation results, a rejection schema is presented to reject those propagation results with low confidence scores. - After the optimal solution is obtained as described above, the system calculates a confidence score for each face as follows:
-
- Taking into account the most similar face in the same photograph that featured ƒi is located in, the confidence score actually reflects the uniqueness of face ƒi in the photo. If the condition Confi<Treject, a predetermined threshold for rejection, is true, then the
name 108 will not be propagated to this face, and instead, the name is kept only loosely associated with an image instead of a face, that is the name is “dangled” with the image. Dangling names are still useful for both photograph searching and further face annotation. For a photographic search, because the names have been associated with photographs, it is straightforward to find these photographs by the associated names. For face annotation in viewer mode 304, if there is a name associated with a photograph that has a face that the user wants to label, the candidate name list generation will be more accurate, as the dangling name is actually a strong prior. - Given both viewer mode 304 and browsing mode 302, users frequently and smoothly alternate between these two modes. Taking into account additional inputs, such as dangling names or labeled faces, face annotation could be made even more efficient, as described next.
- From Annotation to Propagation
- For the
name propagation engine 208 in browsing mode 302, there may be some additional inputs from viewer mode 304, i.e. labeled faces in some images. For example, among the photographs selected by a user, there may be some previously labeled faces, e.g., in the labeled faceslist 220. In this case, the iterative optimization performed by thesimilarity measure engine 232 is adapted to utilize such additional information. - Let P+ be the positive set of faces labeled with the same name as , and let P− be the negative set of faces labeled with different name to . Without losing generality, either P+ or P− can be empty. The iterative optimization performed by the
similarity measure engine 232 in the six steps above is modified according to the following rules: - 1. Faces in P+ are actually part of the solution in Θ={ƒ1, ƒ2, . . . , ƒN}. The influence of these faces is allowed, and only faces Θ\P+ in the corresponding photos in the above Equation (4) are changed.
- 2. Equation (4) is modified to include the influence from P−.
-
- 3. Equation (5) is also modified to include the influence from P−.
-
- With these modifications, the optimal solution will be close to P+ while simultaneously far away from P−.
- From Propagation to Annotation
- For generation of the labeled faces
list 220 in viewer mode 304, there are some additional inputs from the browsing mode 302, i.e., the names dangled at the image level. For example, in a photo in which a user desires to label a face, aname 108 has been associated with the photograph, yet has not been propagated to a face due to low confidence. To utilize dangling names, theview mode manager 236 can adopt an ad hoc strategy by adjusting the prior of dangling name P in photographs Ii as follows: -
- where p(P) is the prior of individual P estimated from historical labeling results,
-
- is actually the prior of the individual P that will appear in this photograph, given that |Fi| is the number of faces in photographs Ii. In this way, the dangling name will be moved forward in the candidate name list and annotation accuracy is improved.
- For example, as shown in
FIG. 7 , the prior of “Peter” is estimated to amount to only 10% from previous labeling results and one name “Peter” is dangling on photograph (a) 702 inFIG. 7 . Given this photograph (a) 702 and the dangling name, the prior that face A is “Peter” is 33%. So, 33% is taken as the prior probability of “Peter” in generating the candidate name list for unknown face A. More importantly, multiple dangling names are also useful for annotation, because any face being annotated reduces uncertainties in the corresponding photograph. For example, in photograph (b) 704 inFIG. 7 , there are two dangling names associated with photograph (b) 704. Once face A is annotated as “Peter”, the name to be associated with face B is immediately determined to be “Tom”. -
FIG. 8 depicts anexemplary method 800 to automatically annotate faces in an image collection. In the flow diagram ofmethod 800, the operations are summarized in individual blocks. The operations may be performed by hardware, software, or combinations of both, for example, by components of an exemplaryface annotation engine 102. - At
block 802, digital images are selected. For example, a browsing mode implemented on a computing device may display thumbnail images of photos stored in a folder on the hard drive. From the entire array of displayed images, a user can select images at will. According to theexemplary method 800, the user selects images that have one or more people in each selected image. The selection itself chooses a set of images that each include a person whose face will be annotated across the selected photos. - At
block 804, the set of images is then named with the name or other moniker of the person common to all the selected images. At this point, the exemplary method does not know whose face in the selected photographs should be associated with the input name. - At
block 806, the name is propagated to a face of the individual represented in each of the images. To accomplish the name propagation, the exemplary method performs a similarity measure function using the features of pairs of faces in the different pictures. The similarity measure can be greatly enhanced by adding contextual features to the faces, e.g., a person's clothes. Computer subdirectory folders of downloaded photos from a digital camera are often limited to a group of photos that were all taken within a short time period, for example at one event. Thus, the human subjects of the photos often wear the same clothes. This extra-facial information enhances similarity comparisons between face pairs. - In one implementation, the exemplary method includes defining an objective function as a sum of similarities between each pair of faces of the individual in different selected images, and maximizing the objective function in order to associate the name with the face in each image. Once an optimal solution is calculated, the name is applied to the face as it is represented in each selected photo. The annotated faces and names are stored to facilitate future name propagation and face annotation.
- The above described systems and methods for propagating face annotations can be implemented in hardware, software, firmware, etc., or combination thereof. In certain implementations, the subject matter may be described in the general context of computer-executable instructions, such as program modules, being executed by a computing device or communications device. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The subject matter can also be practiced in distributed communications environments where tasks are performed over wireless communication by remote processing devices that are linked through a communications network. In a wireless network, program modules may be located in both local and remote communications device storage media including memory storage devices.
- The foregoing discussion describes exemplary systems, engines, user interfaces, and methods for efficient propagation for face annotation. Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features acts described above are disclosed as example forms of implementing the claims.
Claims (20)
1. A computer-executable method, comprising:
receiving a first input designating a collection of digital images;
receiving a second input assigning a name to the collection;
automatically finding a visual target in each of the digital images that the digital images have in common, wherein the visual target in each image comprises a face of an individual person and a visual context near the face; and
propagating the name to the face in each of the digital images.
2. The computer-executable method as recited in claim 1 , wherein the face presents at different angles along a frontal to side-view spectrum of visual perspectives across the digital images.
3. The computer-executable method as recited in claim 1 , wherein at least some of the digital images include multiple different faces.
4. The computer-executable method as recited in claim 1 , further comprising finding the face in only a majority of the digital images.
5. The computer-executable method as recited in claim 1 , wherein the propagating includes:
defining a sum of similarities between each pair of faces across different digital images as an objective function; and
maximizing the objective function to associate the name with the face in each digital image that contains the face.
6. The computer-executable method as recited in claim 1 , wherein the propagating includes:
defining a sum of similarities between each pair of faces across different digital images as an objective function, wherein each pair of faces includes a corresponding visual context near each face; and
maximizing the objective function to associate the name with the face in each digital image that contains the face.
7. The computer-executable method as recited in claim 1 , further comprising improving a name propagation accuracy by using at least some previously named faces of individuals represented in the selected digital images.
8. The computer-executable method as recited in claim 1 , further comprising improving an annotation accuracy of associating the name with the face by using at least some digital images that have been previously associated with names.
9. The computer-executable method as recited in claim 1 , further comprising sharing propagated names information between a browsing mode for viewing and selecting multiple of the digital images and a viewer mode for viewing single digital images and annotating a face in the single digital image.
10. The computer-executable method as recited in claim 9 , wherein the sharing includes sharing a named face from the viewer mode to the browsing mode.
11. The computer-executable method as recited in claim 10 , wherein the sharing includes associating names from the browsing mode with a menu of names for annotating a face in the viewer mode.
12. A storage medium comprising a plurality of executable instructions which, when executed, implement a computer-executable method according to claim 1 .
13. A face annotation engine, comprising:
a selection engine for selecting multiple images, each image having at least a face of an individual person in common, wherein the multiple images are capable of including faces of multiple persons;
a user interface for applying a name to the selected multiple images; and
a name propagation engine for determining the face common to the selected multiple images based at least in part on a visual context near the face and for annotating the name to the face.
14. The face annotation engine as recited in claim 13 , further comprising a labeled faces list for use by a browsing mode manager and a viewer mode manager, wherein the browsing mode manager performs selection and annotation of a group of images and the viewer mode manager performs annotation of a face in a single image.
15. The face annotation engine as recited in claim 13 , wherein the name propagation engine includes a similarity measure engine to define an objective function as a sum of similarities between multiple features of each pair of faces and corresponding visual contexts in different selected images and to maximize the objective function in order to associate the name with the face in each image.
16. The face annotation engine as recited in claim 15 , wherein the name propagation engine includes a contextual features engine to associate multiple faces in different images with the same person based on a non-facial feature that is similar in the different images.
17. The face annotation engine as recited in claim 11 , further comprising a list of previously input names and a menu generator to provide a menu of candidate names for annotating a face in an image.
18. The face annotation engine as recited in claim 11 , further comprising:
a similar face retriever to allow a user to search for similar faces by specifying either a face or a name and to annotate multiple faces in a batch manner.
19. The face annotation engine as recited in claim 11 , wherein the name propagation engine improves a name propagation accuracy by using at least some previously named faces or by using at least some images that have been previously associated with names.
20. A system, comprising:
means for selecting a batch of digital images, wherein each image in the batch has at least a face of one person in common;
means for providing a name for the batch;
means for identifying a face between multiple representations of the face in different images based on at least a feature of the face and a feature of a visual context near the face; and
means for propagating the name to the face in each digital image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US12/047,247 US20080152201A1 (en) | 2005-04-21 | 2008-03-12 | Efficient Propagation for Face Annotation |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/111,696 US7403642B2 (en) | 2005-04-21 | 2005-04-21 | Efficient propagation for face annotation |
US12/047,247 US20080152201A1 (en) | 2005-04-21 | 2008-03-12 | Efficient Propagation for Face Annotation |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/111,696 Continuation US7403642B2 (en) | 2005-04-21 | 2005-04-21 | Efficient propagation for face annotation |
Publications (1)
Publication Number | Publication Date |
---|---|
US20080152201A1 true US20080152201A1 (en) | 2008-06-26 |
Family
ID=37186943
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/111,696 Expired - Fee Related US7403642B2 (en) | 2005-04-21 | 2005-04-21 | Efficient propagation for face annotation |
US12/047,247 Abandoned US20080152201A1 (en) | 2005-04-21 | 2008-03-12 | Efficient Propagation for Face Annotation |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/111,696 Expired - Fee Related US7403642B2 (en) | 2005-04-21 | 2005-04-21 | Efficient propagation for face annotation |
Country Status (1)
Country | Link |
---|---|
US (2) | US7403642B2 (en) |
Cited By (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050280719A1 (en) * | 2004-04-21 | 2005-12-22 | Samsung Electronics Co., Ltd. | Method, medium, and apparatus for detecting situation change of digital photo and method, medium, and apparatus for situation-based photo clustering in digital photo album |
US20090006484A1 (en) * | 2007-06-29 | 2009-01-01 | Nokia Corporation | Apparatus, method and computer program product for using images in contact lists maintained in electronic devices |
US20110291916A1 (en) * | 2008-12-10 | 2011-12-01 | Frank Arndt | Method for transmitting an image from a first control unit to a second control unit and output unit |
US8532387B2 (en) | 2009-09-04 | 2013-09-10 | Adobe Systems Incorporated | Methods and apparatus for procedural directional texture generation |
US20130308836A1 (en) * | 2012-05-18 | 2013-11-21 | Primax Electronics Ltd. | Photo image managing method and photo image managing system |
US8599219B2 (en) | 2009-09-18 | 2013-12-03 | Adobe Systems Incorporated | Methods and apparatuses for generating thumbnail summaries for image collections |
US8619098B2 (en) * | 2009-09-18 | 2013-12-31 | Adobe Systems Incorporated | Methods and apparatuses for generating co-salient thumbnails for digital images |
US20140233811A1 (en) * | 2012-05-15 | 2014-08-21 | Google Inc. | Summarizing a photo album |
US20140301653A1 (en) * | 2012-05-15 | 2014-10-09 | Google Inc. | Summarizing a photo album in a social network system |
US8861868B2 (en) | 2011-08-29 | 2014-10-14 | Adobe-Systems Incorporated | Patch-based synthesis techniques |
US9857941B2 (en) | 2010-01-06 | 2018-01-02 | Apple Inc. | Device, method, and graphical user interface for navigating and displaying content in context |
DK201670608A1 (en) * | 2016-06-12 | 2018-01-02 | Apple Inc | User interfaces for retrieving contextually relevant media content |
US9858295B2 (en) | 2014-06-24 | 2018-01-02 | Google Llc | Ranking and selecting images for display from a set of images |
US10324973B2 (en) | 2016-06-12 | 2019-06-18 | Apple Inc. | Knowledge graph metadata network based on notable moments |
US10362219B2 (en) | 2016-09-23 | 2019-07-23 | Apple Inc. | Avatar creation and editing |
US10572132B2 (en) | 2015-06-05 | 2020-02-25 | Apple Inc. | Formatting content for a reduced-size user interface |
US10739974B2 (en) | 2016-06-11 | 2020-08-11 | Apple Inc. | Configuring context-specific user interfaces |
US10803135B2 (en) | 2018-09-11 | 2020-10-13 | Apple Inc. | Techniques for disambiguating clustered occurrence identifiers |
US10846343B2 (en) | 2018-09-11 | 2020-11-24 | Apple Inc. | Techniques for disambiguating clustered location identifiers |
US10904426B2 (en) | 2006-09-06 | 2021-01-26 | Apple Inc. | Portable electronic device for photo management |
US11086935B2 (en) | 2018-05-07 | 2021-08-10 | Apple Inc. | Smart updates from historical database changes |
US11157138B2 (en) * | 2017-05-31 | 2021-10-26 | International Business Machines Corporation | Thumbnail generation for digital images |
US11243996B2 (en) | 2018-05-07 | 2022-02-08 | Apple Inc. | Digital asset search user interface |
US11307737B2 (en) | 2019-05-06 | 2022-04-19 | Apple Inc. | Media browsing user interface with intelligently selected representative media items |
US11334209B2 (en) * | 2016-06-12 | 2022-05-17 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US11782575B2 (en) | 2018-05-07 | 2023-10-10 | Apple Inc. | User interfaces for sharing contextually relevant media content |
US11854539B2 (en) | 2018-05-07 | 2023-12-26 | Apple Inc. | Intelligent automated assistant for delivering content from user experiences |
Families Citing this family (110)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8189927B2 (en) * | 2007-03-05 | 2012-05-29 | DigitalOptics Corporation Europe Limited | Face categorization and annotation of a mobile phone contact list |
US8553949B2 (en) | 2004-01-22 | 2013-10-08 | DigitalOptics Corporation Europe Limited | Classification and organization of consumer digital images using workflow, and face detection and recognition |
US8363951B2 (en) * | 2007-03-05 | 2013-01-29 | DigitalOptics Corporation Europe Limited | Face recognition training method and apparatus |
US7564994B1 (en) | 2004-01-22 | 2009-07-21 | Fotonation Vision Limited | Classification system for consumer digital images using automatic workflow and face detection and recognition |
US8563133B2 (en) * | 2004-06-08 | 2013-10-22 | Sandisk Corporation | Compositions and methods for modulation of nanostructure energy levels |
US9405751B2 (en) | 2005-08-23 | 2016-08-02 | Ricoh Co., Ltd. | Database for mixed media document system |
US8868555B2 (en) | 2006-07-31 | 2014-10-21 | Ricoh Co., Ltd. | Computation of a recongnizability score (quality predictor) for image retrieval |
US8965145B2 (en) | 2006-07-31 | 2015-02-24 | Ricoh Co., Ltd. | Mixed media reality recognition using multiple specialized indexes |
US9171202B2 (en) | 2005-08-23 | 2015-10-27 | Ricoh Co., Ltd. | Data organization and access for mixed media document system |
US8856108B2 (en) | 2006-07-31 | 2014-10-07 | Ricoh Co., Ltd. | Combining results of image retrieval processes |
US8838591B2 (en) * | 2005-08-23 | 2014-09-16 | Ricoh Co., Ltd. | Embedding hot spots in electronic documents |
US7812986B2 (en) | 2005-08-23 | 2010-10-12 | Ricoh Co. Ltd. | System and methods for use of voice mail and email in a mixed media environment |
US8600989B2 (en) | 2004-10-01 | 2013-12-03 | Ricoh Co., Ltd. | Method and system for image matching in a mixed media environment |
US8521737B2 (en) * | 2004-10-01 | 2013-08-27 | Ricoh Co., Ltd. | Method and system for multi-tier image matching in a mixed media environment |
US9530050B1 (en) | 2007-07-11 | 2016-12-27 | Ricoh Co., Ltd. | Document annotation sharing |
US8510283B2 (en) | 2006-07-31 | 2013-08-13 | Ricoh Co., Ltd. | Automatic adaption of an image recognition system to image capture devices |
US8825682B2 (en) | 2006-07-31 | 2014-09-02 | Ricoh Co., Ltd. | Architecture for mixed media reality retrieval of locations and registration of images |
US8156116B2 (en) | 2006-07-31 | 2012-04-10 | Ricoh Co., Ltd | Dynamic presentation of targeted information in a mixed media reality recognition system |
US8176054B2 (en) | 2007-07-12 | 2012-05-08 | Ricoh Co. Ltd | Retrieving electronic documents by converting them to synthetic text |
US9384619B2 (en) | 2006-07-31 | 2016-07-05 | Ricoh Co., Ltd. | Searching media content for objects specified using identifiers |
US8156115B1 (en) | 2007-07-11 | 2012-04-10 | Ricoh Co. Ltd. | Document-based networking with mixed media reality |
US9373029B2 (en) | 2007-07-11 | 2016-06-21 | Ricoh Co., Ltd. | Invisible junction feature recognition for document security or annotation |
US8949287B2 (en) | 2005-08-23 | 2015-02-03 | Ricoh Co., Ltd. | Embedding hot spots in imaged documents |
US7702673B2 (en) | 2004-10-01 | 2010-04-20 | Ricoh Co., Ltd. | System and methods for creation and use of a mixed media environment |
US7715597B2 (en) | 2004-12-29 | 2010-05-11 | Fotonation Ireland Limited | Method and component for image recognition |
US7590267B2 (en) * | 2005-05-31 | 2009-09-15 | Microsoft Corporation | Accelerated face detection based on prior probability of a view |
JP2007080184A (en) * | 2005-09-16 | 2007-03-29 | Canon Inc | Image processor and method |
US20070098303A1 (en) * | 2005-10-31 | 2007-05-03 | Eastman Kodak Company | Determining a particular person from a collection |
US7945653B2 (en) * | 2006-10-11 | 2011-05-17 | Facebook, Inc. | Tagging digital media |
US7720258B1 (en) | 2006-01-26 | 2010-05-18 | Adobe Systems Incorporated | Structured comparison of objects from similar images |
US7813526B1 (en) | 2006-01-26 | 2010-10-12 | Adobe Systems Incorporated | Normalizing detected objects |
US7716157B1 (en) | 2006-01-26 | 2010-05-11 | Adobe Systems Incorporated | Searching images with extracted objects |
US7636450B1 (en) | 2006-01-26 | 2009-12-22 | Adobe Systems Incorporated | Displaying detected objects to indicate grouping |
US7813557B1 (en) | 2006-01-26 | 2010-10-12 | Adobe Systems Incorporated | Tagging detected objects |
US7694885B1 (en) | 2006-01-26 | 2010-04-13 | Adobe Systems Incorporated | Indicating a tag with visual data |
US8259995B1 (en) | 2006-01-26 | 2012-09-04 | Adobe Systems Incorporated | Designating a tag icon |
US7706577B1 (en) | 2006-01-26 | 2010-04-27 | Adobe Systems Incorporated | Exporting extracted faces |
US7978936B1 (en) * | 2006-01-26 | 2011-07-12 | Adobe Systems Incorporated | Indicating a correspondence between an image and an object |
US8024343B2 (en) | 2006-04-07 | 2011-09-20 | Eastman Kodak Company | Identifying unique objects in multiple image collections |
US8065313B2 (en) * | 2006-07-24 | 2011-11-22 | Google Inc. | Method and apparatus for automatically annotating images |
US8489987B2 (en) | 2006-07-31 | 2013-07-16 | Ricoh Co., Ltd. | Monitoring and analyzing creation and usage of visual content using image and hotspot interaction |
US9063952B2 (en) | 2006-07-31 | 2015-06-23 | Ricoh Co., Ltd. | Mixed media reality recognition with image tracking |
US9020966B2 (en) | 2006-07-31 | 2015-04-28 | Ricoh Co., Ltd. | Client device for interacting with a mixed media reality recognition system |
US8676810B2 (en) | 2006-07-31 | 2014-03-18 | Ricoh Co., Ltd. | Multiple index mixed media reality recognition using unequal priority indexes |
US9176984B2 (en) | 2006-07-31 | 2015-11-03 | Ricoh Co., Ltd | Mixed media reality retrieval of differentially-weighted links |
US8201076B2 (en) | 2006-07-31 | 2012-06-12 | Ricoh Co., Ltd. | Capturing symbolic information from documents upon printing |
US7515740B2 (en) | 2006-08-02 | 2009-04-07 | Fotonation Vision Limited | Face recognition with combined PCA-based datasets |
US20080077595A1 (en) * | 2006-09-14 | 2008-03-27 | Eric Leebow | System and method for facilitating online social networking |
US8436911B2 (en) * | 2006-09-14 | 2013-05-07 | Freezecrowd, Inc. | Tagging camera |
US7916976B1 (en) | 2006-10-05 | 2011-03-29 | Kedikian Roland H | Facial based image organization and retrieval method |
US8315463B2 (en) * | 2006-11-14 | 2012-11-20 | Eastman Kodak Company | User interface for face recognition |
US8276098B2 (en) | 2006-12-22 | 2012-09-25 | Apple Inc. | Interactive image thumbnails |
US9142253B2 (en) * | 2006-12-22 | 2015-09-22 | Apple Inc. | Associating keywords to media |
KR100796044B1 (en) * | 2007-02-08 | 2008-01-21 | (주)올라웍스 | Method for tagging a person image |
US8014572B2 (en) * | 2007-06-08 | 2011-09-06 | Microsoft Corporation | Face annotation framework with partial clustering and interactive labeling |
US7996762B2 (en) * | 2007-09-21 | 2011-08-09 | Microsoft Corporation | Correlative multi-label image annotation |
CN101398832A (en) * | 2007-09-30 | 2009-04-01 | 国际商业机器公司 | Image searching method and system by utilizing human face detection |
CA2711143C (en) * | 2007-12-31 | 2015-12-08 | Ray Ganong | Method, system, and computer program for identification and sharing of digital images with face signatures |
US9639740B2 (en) | 2007-12-31 | 2017-05-02 | Applied Recognition Inc. | Face detection and recognition |
US9721148B2 (en) | 2007-12-31 | 2017-08-01 | Applied Recognition Inc. | Face detection and recognition |
US20090316961A1 (en) * | 2008-06-21 | 2009-12-24 | Microsoft Corporation | Method for tagging image content |
US8339451B2 (en) * | 2008-06-27 | 2012-12-25 | Nokia Corporation | Image navigation with multiple images |
JP5371083B2 (en) * | 2008-09-16 | 2013-12-18 | Kddi株式会社 | Face identification feature value registration apparatus, face identification feature value registration method, face identification feature value registration program, and recording medium |
WO2010063463A2 (en) | 2008-12-05 | 2010-06-10 | Fotonation Ireland Limited | Face recognition using face tracker classifier data |
US9727312B1 (en) * | 2009-02-17 | 2017-08-08 | Ikorongo Technology, LLC | Providing subject information regarding upcoming images on a display |
US8121358B2 (en) * | 2009-03-06 | 2012-02-21 | Cyberlink Corp. | Method of grouping images by face |
US8531478B2 (en) * | 2009-03-19 | 2013-09-10 | Cyberlink Corp. | Method of browsing photos based on people |
WO2010125781A1 (en) * | 2009-04-27 | 2010-11-04 | パナソニック株式会社 | Data processing device, data processing method, program, and integrated circuit |
US8325999B2 (en) * | 2009-06-08 | 2012-12-04 | Microsoft Corporation | Assisted face recognition tagging |
US9710491B2 (en) * | 2009-11-02 | 2017-07-18 | Microsoft Technology Licensing, Llc | Content-based image search |
US20110106798A1 (en) * | 2009-11-02 | 2011-05-05 | Microsoft Corporation | Search Result Enhancement Through Image Duplicate Detection |
US8433140B2 (en) * | 2009-11-02 | 2013-04-30 | Microsoft Corporation | Image metadata propagation |
US8416997B2 (en) * | 2010-01-27 | 2013-04-09 | Apple Inc. | Method of person identification using social connections |
US8645287B2 (en) | 2010-02-04 | 2014-02-04 | Microsoft Corporation | Image tagging based upon cross domain context |
CN102202173B (en) * | 2010-03-23 | 2013-01-16 | 三星电子(中国)研发中心 | Photo automatically naming method and device thereof |
KR20120021061A (en) * | 2010-08-31 | 2012-03-08 | 삼성전자주식회사 | Method for providing search service to extract keywords in specific region and display apparatus applying the same |
JP2012073998A (en) * | 2010-08-31 | 2012-04-12 | Casio Comput Co Ltd | Image distribution system, image display device, image distribution server, and program |
KR20120021057A (en) * | 2010-08-31 | 2012-03-08 | 삼성전자주식회사 | Method for providing search service to extract keywords in specific region and display apparatus applying the same |
US8724908B2 (en) | 2010-09-03 | 2014-05-13 | Adobe Systems Incorporated | System and method for labeling a collection of images |
JP5134664B2 (en) * | 2010-09-14 | 2013-01-30 | 株式会社東芝 | Annotation device |
US9552376B2 (en) | 2011-06-09 | 2017-01-24 | MemoryWeb, LLC | Method and apparatus for managing digital files |
US20120314916A1 (en) * | 2011-06-13 | 2012-12-13 | Reagan Inventions, Llc | Identifying and tagging objects within a digital image |
US8582828B2 (en) * | 2011-06-24 | 2013-11-12 | Google Inc. | Using photographs to manage groups |
US9058331B2 (en) | 2011-07-27 | 2015-06-16 | Ricoh Co., Ltd. | Generating a conversation in a social network based on visual search results |
US8548207B2 (en) | 2011-08-15 | 2013-10-01 | Daon Holdings Limited | Method of host-directed illumination and system for conducting host-directed illumination |
CN103064864A (en) * | 2011-10-19 | 2013-04-24 | 致伸科技股份有限公司 | Photo sharing system with face recognition function |
US9256620B2 (en) * | 2011-12-20 | 2016-02-09 | Amazon Technologies, Inc. | Techniques for grouping images |
US9202105B1 (en) | 2012-01-13 | 2015-12-01 | Amazon Technologies, Inc. | Image analysis for user authentication |
US8838432B2 (en) * | 2012-02-06 | 2014-09-16 | Microsoft Corporation | Image annotations on web pages |
US9239848B2 (en) | 2012-02-06 | 2016-01-19 | Microsoft Technology Licensing, Llc | System and method for semantically annotating images |
US9116924B2 (en) | 2013-01-14 | 2015-08-25 | Xerox Corporation | System and method for image selection using multivariate time series analysis |
US11250203B2 (en) * | 2013-08-12 | 2022-02-15 | Microsoft Technology Licensing, Llc | Browsing images via mined hyperlinked text snippets |
US10540541B2 (en) * | 2014-05-27 | 2020-01-21 | International Business Machines Corporation | Cognitive image detection and recognition |
JP6215774B2 (en) | 2014-05-29 | 2017-10-18 | 富士フイルム株式会社 | Same person determination device and method and control program therefor |
US9195912B1 (en) * | 2014-07-24 | 2015-11-24 | National Taipei University Of Technology | Face annotation method and a face annotation system |
CA3186147A1 (en) | 2014-08-28 | 2016-02-28 | Kevin Alan Tussy | Facial recognition authentication system including path parameters |
US11256792B2 (en) | 2014-08-28 | 2022-02-22 | Facetec, Inc. | Method and apparatus for creation and use of digital identification |
US10803160B2 (en) | 2014-08-28 | 2020-10-13 | Facetec, Inc. | Method to verify and identify blockchain with user question data |
US10614204B2 (en) | 2014-08-28 | 2020-04-07 | Facetec, Inc. | Facial recognition authentication system including path parameters |
US10915618B2 (en) | 2014-08-28 | 2021-02-09 | Facetec, Inc. | Method to add remotely collected biometric images / templates to a database record of personal information |
US10698995B2 (en) | 2014-08-28 | 2020-06-30 | Facetec, Inc. | Method to verify identity using a previously collected biometric image/data |
US9747648B2 (en) * | 2015-01-20 | 2017-08-29 | Kuo-Chun Fang | Systems and methods for publishing data on social media websites |
CN106155508B (en) * | 2015-04-01 | 2021-01-01 | 腾讯科技(上海)有限公司 | Information processing method and client |
JP2017033547A (en) * | 2015-08-05 | 2017-02-09 | キヤノン株式会社 | Information processing apparatus, control method therefor, and program |
CN105354307B (en) * | 2015-11-06 | 2021-01-15 | 腾讯科技(深圳)有限公司 | Image content identification method and device |
USD987653S1 (en) | 2016-04-26 | 2023-05-30 | Facetec, Inc. | Display screen or portion thereof with graphical user interface |
US11290530B2 (en) | 2018-06-01 | 2022-03-29 | Apple Inc. | Customizable, pull-based asset transfer requests using object models |
CN109033220B (en) * | 2018-06-29 | 2022-09-06 | 北京京东尚科信息技术有限公司 | Automatic selection method, system, equipment and storage medium of labeled data |
CN110147458A (en) * | 2019-05-24 | 2019-08-20 | 涂哲 | A kind of photo screening technique, system and electric terminal |
EP4010845A1 (en) * | 2019-08-09 | 2022-06-15 | Clearview AI, Inc. | Methods for providing information about a person based on facial recognition |
Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5781650A (en) * | 1994-02-18 | 1998-07-14 | University Of Central Florida | Automatic feature detection and age classification of human faces in digital images |
US6445810B2 (en) * | 1997-08-01 | 2002-09-03 | Interval Research Corporation | Method and apparatus for personnel detection and tracking |
US20030210808A1 (en) * | 2002-05-10 | 2003-11-13 | Eastman Kodak Company | Method and apparatus for organizing and retrieving images containing human faces |
US20040085578A1 (en) * | 2002-11-03 | 2004-05-06 | Quek Su Mien | Producing personalized photo calendar |
US20040122539A1 (en) * | 2002-12-20 | 2004-06-24 | Ainsworth Heather C. | Synchronization of music and images in a digital multimedia device system |
US20040145602A1 (en) * | 2003-01-24 | 2004-07-29 | Microsoft Corporation | Organizing and displaying photographs based on time |
US6785421B1 (en) * | 2000-05-22 | 2004-08-31 | Eastman Kodak Company | Analyzing images to determine if one or more sets of materials correspond to the analyzed images |
US20040213553A1 (en) * | 2003-01-29 | 2004-10-28 | Seiko Epson Corporation | Image retrieving device, method for adding keywords in image retrieving device, and computer program therefor |
US20040247177A1 (en) * | 2003-06-05 | 2004-12-09 | Canon Kabushiki Kaisha | Image processing |
US20040264780A1 (en) * | 2003-06-30 | 2004-12-30 | Lei Zhang | Face annotation for photo management |
US6940545B1 (en) * | 2000-02-28 | 2005-09-06 | Eastman Kodak Company | Face detecting camera and method |
US20060071947A1 (en) * | 2004-10-06 | 2006-04-06 | Randy Ubillos | Techniques for displaying digital images on a display |
US20060140455A1 (en) * | 2004-12-29 | 2006-06-29 | Gabriel Costache | Method and component for image recognition |
US20070043459A1 (en) * | 1999-12-15 | 2007-02-22 | Tangis Corporation | Storing and recalling information to augment human memories |
US20070230799A1 (en) * | 2000-08-29 | 2007-10-04 | Imageid Ltd. | Indexing, storage and retrieval of digital images |
US7327891B2 (en) * | 2001-07-17 | 2008-02-05 | Yesvideo, Inc. | Automatic selection of a visual image or images from a collection of visual images, based on an evaluation of the quality of the visual images |
US7362919B2 (en) * | 2002-12-12 | 2008-04-22 | Eastman Kodak Company | Method for generating customized photo album pages and prints based on people and gender profiles |
-
2005
- 2005-04-21 US US11/111,696 patent/US7403642B2/en not_active Expired - Fee Related
-
2008
- 2008-03-12 US US12/047,247 patent/US20080152201A1/en not_active Abandoned
Patent Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5781650A (en) * | 1994-02-18 | 1998-07-14 | University Of Central Florida | Automatic feature detection and age classification of human faces in digital images |
US6445810B2 (en) * | 1997-08-01 | 2002-09-03 | Interval Research Corporation | Method and apparatus for personnel detection and tracking |
US20070043459A1 (en) * | 1999-12-15 | 2007-02-22 | Tangis Corporation | Storing and recalling information to augment human memories |
US6940545B1 (en) * | 2000-02-28 | 2005-09-06 | Eastman Kodak Company | Face detecting camera and method |
US6785421B1 (en) * | 2000-05-22 | 2004-08-31 | Eastman Kodak Company | Analyzing images to determine if one or more sets of materials correspond to the analyzed images |
US20070230799A1 (en) * | 2000-08-29 | 2007-10-04 | Imageid Ltd. | Indexing, storage and retrieval of digital images |
US7327891B2 (en) * | 2001-07-17 | 2008-02-05 | Yesvideo, Inc. | Automatic selection of a visual image or images from a collection of visual images, based on an evaluation of the quality of the visual images |
US20030210808A1 (en) * | 2002-05-10 | 2003-11-13 | Eastman Kodak Company | Method and apparatus for organizing and retrieving images containing human faces |
US20040085578A1 (en) * | 2002-11-03 | 2004-05-06 | Quek Su Mien | Producing personalized photo calendar |
US7362919B2 (en) * | 2002-12-12 | 2008-04-22 | Eastman Kodak Company | Method for generating customized photo album pages and prints based on people and gender profiles |
US20040122539A1 (en) * | 2002-12-20 | 2004-06-24 | Ainsworth Heather C. | Synchronization of music and images in a digital multimedia device system |
US20040145602A1 (en) * | 2003-01-24 | 2004-07-29 | Microsoft Corporation | Organizing and displaying photographs based on time |
US20040213553A1 (en) * | 2003-01-29 | 2004-10-28 | Seiko Epson Corporation | Image retrieving device, method for adding keywords in image retrieving device, and computer program therefor |
US20040247177A1 (en) * | 2003-06-05 | 2004-12-09 | Canon Kabushiki Kaisha | Image processing |
US20040264780A1 (en) * | 2003-06-30 | 2004-12-30 | Lei Zhang | Face annotation for photo management |
US20060071947A1 (en) * | 2004-10-06 | 2006-04-06 | Randy Ubillos | Techniques for displaying digital images on a display |
US20060140455A1 (en) * | 2004-12-29 | 2006-06-29 | Gabriel Costache | Method and component for image recognition |
Cited By (51)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050280719A1 (en) * | 2004-04-21 | 2005-12-22 | Samsung Electronics Co., Ltd. | Method, medium, and apparatus for detecting situation change of digital photo and method, medium, and apparatus for situation-based photo clustering in digital photo album |
US10904426B2 (en) | 2006-09-06 | 2021-01-26 | Apple Inc. | Portable electronic device for photo management |
US11601584B2 (en) | 2006-09-06 | 2023-03-07 | Apple Inc. | Portable electronic device for photo management |
US9571675B2 (en) * | 2007-06-29 | 2017-02-14 | Nokia Technologies Oy | Apparatus, method and computer program product for using images in contact lists maintained in electronic devices |
US20090006484A1 (en) * | 2007-06-29 | 2009-01-01 | Nokia Corporation | Apparatus, method and computer program product for using images in contact lists maintained in electronic devices |
US20110291916A1 (en) * | 2008-12-10 | 2011-12-01 | Frank Arndt | Method for transmitting an image from a first control unit to a second control unit and output unit |
US8532387B2 (en) | 2009-09-04 | 2013-09-10 | Adobe Systems Incorporated | Methods and apparatus for procedural directional texture generation |
US8787698B2 (en) | 2009-09-04 | 2014-07-22 | Adobe Systems Incorporated | Methods and apparatus for directional texture generation using image warping |
US8599219B2 (en) | 2009-09-18 | 2013-12-03 | Adobe Systems Incorporated | Methods and apparatuses for generating thumbnail summaries for image collections |
US8619098B2 (en) * | 2009-09-18 | 2013-12-31 | Adobe Systems Incorporated | Methods and apparatuses for generating co-salient thumbnails for digital images |
US11592959B2 (en) | 2010-01-06 | 2023-02-28 | Apple Inc. | Device, method, and graphical user interface for navigating and displaying content in context |
US9857941B2 (en) | 2010-01-06 | 2018-01-02 | Apple Inc. | Device, method, and graphical user interface for navigating and displaying content in context |
US11099712B2 (en) | 2010-01-06 | 2021-08-24 | Apple Inc. | Device, method, and graphical user interface for navigating and displaying content in context |
US10296166B2 (en) | 2010-01-06 | 2019-05-21 | Apple Inc. | Device, method, and graphical user interface for navigating and displaying content in context |
US10732790B2 (en) | 2010-01-06 | 2020-08-04 | Apple Inc. | Device, method, and graphical user interface for navigating and displaying content in context |
US8861868B2 (en) | 2011-08-29 | 2014-10-14 | Adobe-Systems Incorporated | Patch-based synthesis techniques |
US9317773B2 (en) | 2011-08-29 | 2016-04-19 | Adobe Systems Incorporated | Patch-based synthesis techniques using color and color gradient voting |
US20140301653A1 (en) * | 2012-05-15 | 2014-10-09 | Google Inc. | Summarizing a photo album in a social network system |
US8891883B2 (en) * | 2012-05-15 | 2014-11-18 | Google Inc. | Summarizing a photo album in a social network system |
US9311530B1 (en) | 2012-05-15 | 2016-04-12 | Google Inc. | Summarizing a photo album in a social network system |
US20140233811A1 (en) * | 2012-05-15 | 2014-08-21 | Google Inc. | Summarizing a photo album |
US20130308836A1 (en) * | 2012-05-18 | 2013-11-21 | Primax Electronics Ltd. | Photo image managing method and photo image managing system |
US9858295B2 (en) | 2014-06-24 | 2018-01-02 | Google Llc | Ranking and selecting images for display from a set of images |
US10417277B2 (en) | 2014-06-24 | 2019-09-17 | Google Llc | Ranking and selecting images for display from a set of images |
US10572132B2 (en) | 2015-06-05 | 2020-02-25 | Apple Inc. | Formatting content for a reduced-size user interface |
US11073799B2 (en) | 2016-06-11 | 2021-07-27 | Apple Inc. | Configuring context-specific user interfaces |
US11733656B2 (en) | 2016-06-11 | 2023-08-22 | Apple Inc. | Configuring context-specific user interfaces |
US10739974B2 (en) | 2016-06-11 | 2020-08-11 | Apple Inc. | Configuring context-specific user interfaces |
US11334209B2 (en) * | 2016-06-12 | 2022-05-17 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US20230297206A1 (en) * | 2016-06-12 | 2023-09-21 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US10891013B2 (en) * | 2016-06-12 | 2021-01-12 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US11941223B2 (en) * | 2016-06-12 | 2024-03-26 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US10324973B2 (en) | 2016-06-12 | 2019-06-18 | Apple Inc. | Knowledge graph metadata network based on notable moments |
US11681408B2 (en) * | 2016-06-12 | 2023-06-20 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
DK201670608A1 (en) * | 2016-06-12 | 2018-01-02 | Apple Inc | User interfaces for retrieving contextually relevant media content |
US10073584B2 (en) | 2016-06-12 | 2018-09-11 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US20220276750A1 (en) * | 2016-06-12 | 2022-09-01 | Apple Inc. | User interfaces for retrieving contextually relevant media content |
US10362219B2 (en) | 2016-09-23 | 2019-07-23 | Apple Inc. | Avatar creation and editing |
US11169661B2 (en) | 2017-05-31 | 2021-11-09 | International Business Machines Corporation | Thumbnail generation for digital images |
US11157138B2 (en) * | 2017-05-31 | 2021-10-26 | International Business Machines Corporation | Thumbnail generation for digital images |
US11782575B2 (en) | 2018-05-07 | 2023-10-10 | Apple Inc. | User interfaces for sharing contextually relevant media content |
US11243996B2 (en) | 2018-05-07 | 2022-02-08 | Apple Inc. | Digital asset search user interface |
US11854539B2 (en) | 2018-05-07 | 2023-12-26 | Apple Inc. | Intelligent automated assistant for delivering content from user experiences |
US11900923B2 (en) | 2018-05-07 | 2024-02-13 | Apple Inc. | Intelligent automated assistant for delivering content from user experiences |
US11086935B2 (en) | 2018-05-07 | 2021-08-10 | Apple Inc. | Smart updates from historical database changes |
US10803135B2 (en) | 2018-09-11 | 2020-10-13 | Apple Inc. | Techniques for disambiguating clustered occurrence identifiers |
US11775590B2 (en) | 2018-09-11 | 2023-10-03 | Apple Inc. | Techniques for disambiguating clustered location identifiers |
US10846343B2 (en) | 2018-09-11 | 2020-11-24 | Apple Inc. | Techniques for disambiguating clustered location identifiers |
US11625153B2 (en) | 2019-05-06 | 2023-04-11 | Apple Inc. | Media browsing user interface with intelligently selected representative media items |
US11307737B2 (en) | 2019-05-06 | 2022-04-19 | Apple Inc. | Media browsing user interface with intelligently selected representative media items |
US11947778B2 (en) | 2019-05-06 | 2024-04-02 | Apple Inc. | Media browsing user interface with intelligently selected representative media items |
Also Published As
Publication number | Publication date |
---|---|
US7403642B2 (en) | 2008-07-22 |
US20060239515A1 (en) | 2006-10-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7403642B2 (en) | Efficient propagation for face annotation | |
Zhang et al. | Efficient propagation for face annotation in family albums | |
US9430719B2 (en) | System and method for providing objectified image renderings using recognition information from images | |
US7809722B2 (en) | System and method for enabling search and retrieval from image files based on recognized information | |
US7519200B2 (en) | System and method for enabling the use of captured images through recognition | |
US7274822B2 (en) | Face annotation for photo management | |
US8027549B2 (en) | System and method for searching a multimedia database using a pictorial language | |
EP2232381B1 (en) | Interactive concept learning in image search | |
US8358811B2 (en) | Method and apparatus to incorporate automatic face recognition in digital image collections | |
Platt | AutoAlbum: Clustering digital photographs using probabilistic model merging | |
JP5123288B2 (en) | Form connections between image collections | |
Zhang et al. | Automated annotation of human faces in family albums | |
US6804684B2 (en) | Method for associating semantic information with multiple images in an image database environment | |
US20060251292A1 (en) | System and method for recognizing objects from images and identifying relevancy amongst images and information | |
US20070288453A1 (en) | System and Method for Searching Multimedia using Exemplar Images | |
JP2006511859A (en) | Search for images | |
EP1889207A2 (en) | System and method for enabling the use of captured images through recognition | |
Choi et al. | Face annotation for personal photos using context-assisted face recognition | |
Thiebauta et al. | On-board compression algorithm for satellite multispectral images | |
de Silva et al. | Embedded tags and visual querying for face photo retrieval | |
Mironică et al. | A relevance feedback approach to video genre retrieval | |
Hibino et al. | Semantics Meets UX: Mediating Intelligent Indexing of Consumers' Multimedia Collections for Multifaceted Visualization and Media Creation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |
|
AS | Assignment |
Owner name: MICROSOFT TECHNOLOGY LICENSING, LLC, WASHINGTON Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MICROSOFT CORPORATION;REEL/FRAME:034766/0001 Effective date: 20141014 |