Methods, systems and apparatus for identifying modified images based on seed images that are known to be modified images. In an aspect, a method includes accessing data identifying a set of first seed images; for each first seed image, determining a respective first set of similar images from images
Methods, systems and apparatus for identifying modified images based on seed images that are known to be modified images. In an aspect, a method includes accessing data identifying a set of first seed images; for each first seed image, determining a respective first set of similar images from images in an image corpus, each similar image having a visual similarity score that is a measure of visual similarity of the similar image to the first seed image based on the image content of the similar image and the first seed image that satisfies a first seed image similarity threshold; and for each similar image in each respective first set of similar images, attributing to the similar image signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold.
대표청구항▼
1. A computer-implemented method, comprising: accessing, by a computer system, data identifying a set of first seed images, each first seed image being classified as belonging to a first category of images based on signal data of the first seed image that are independent of image content of the firs
1. A computer-implemented method, comprising: accessing, by a computer system, data identifying a set of first seed images, each first seed image being classified as belonging to a first category of images based on signal data of the first seed image that are independent of image content of the first seed image, and wherein the set of first seed images is a proper subset of images in an image corpus; for each first seed image:determining, by the computer system, a respective first set of similar images from images in the image corpus, each similar image in the respective first set of images having a visual similarity score that is a measure of visual similarity of the similar image to the first seed image based on the image content of the similar image and the first seed image, and that satisfies a first seed image similarity threshold; andfor each similar image in each respective first set of similar images:attributing, by the computer system, to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold;determining, by the computer system, whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image; andclassifying, by the computer system, only the similar images that are determined to belong the first category as belonging to the first category of images; wherein the first category of images are images that have been classified as spoof images. 2. The computer-implemented method of claim 1, wherein: each seed image is an image for which a classification score generated based on the signal data meets a first classification score threshold that corresponds to a first likelihood that an image belongs to the first category; anddetermining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image comprises: generating a classification score based on the signal data attributed to the similar image; anddetermining that the classification score meets at least a second classification score threshold that corresponds to a second likelihood that an image belongs to the first category, and wherein the second likelihood is less than the first likelihood. 3. The computer-implemented method of claim 1, further comprising: for each first seed image: determining a respective first set of similar seed images from the other first seed images, each similar seed image in the respective first set of similar seed images having a visual similarity score that is a measure of visual similarity of the similar seed image to the first seed image based on the image content of similar image and the first seed image, and that satisfies a first seed image similarity threshold; andwherein, for each similar image in each respective first set of similar images, attributing to the similar image the signal data of each first seed image comprises: attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold, and the signal data attributed to each first seed image from the similar seed images. 4. The computer-implemented method of claim 1, further comprising: accessing data identifying a set of second seed images, each second seed image being classified as not belonging to the first category of images based on signal data of the second seed image that are independent of image content of the second seed image, and wherein the set of second seed images is a proper subset of images in an image corpus;for each second seed image: determining a respective second set of similar images from images in the image corpus, each similar image in the respective second set of images having a visual similarity score that is a measure of visual similarity of the similar image to the second seed image based on the image content of the similar image and the second seed image, and that satisfies a first seed image similarity threshold; andfor each similar image in each respective second set of similar images: attributing to the similar image the signal data of each second seed image for which the similar image has a respective visual similarity score satisfying the second seed image similarity threshold; andwherein determining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image comprises, for each similar image belonging a respective first set and a respective second set, determining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image from the first seed images and the second seed images. 5. The computer-implemented method of claim 1, wherein the signal data of the first seed images comprise click metric data indicative of one or more click metrics for the first seed image. 6. The computer-implemented method of claim 1, wherein the signal data of the first seed images comprise hover metric data indicative of one or more hover metrics for the first seed image. 7. The computer-implemented method of claim 1, wherein the signal data are data indicating an image that has received of a disproportion number of selections relative to a quality measure of the image. 8. The computer-implemented method of claim 1, wherein attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold comprises summing the signal data and associating the summed signal data with the similar image. 9. The computer-implemented method of claim 1, wherein attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold comprises generating a central tendency of the signal data and associating the central tendency of the signal data with the similar image. 10. The computer-implemented method of claim 1, wherein attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold comprises adjusting signal data of the first seed image attributed to the similar image by a value that is proportional to the visual similarity score of the similar image. 11. A system, comprising: a data processing apparatus; anda non-transitory memory storage apparatus in data communication with the data processing apparatus and storing instructions executable by the data processing apparatus and that upon such execution cause the data processing apparatus to perform operations comprising:accessing data identifying a set of first seed images, each first seed image being classified as belonging to a first category of images based on signal data of the first seed image that are independent of image content of the first seed image, and wherein the set of first seed images is a proper subset of images in an image corpus;for each first seed image:determining a respective first set of similar images from images in the image corpus, each similar image in the respective first set of images having a visual similarity score that is a measure of visual similarity of the similar image to the first seed image based on the image content of the similar image and the first seed image, and that satisfies a first seed image similarity threshold; andfor each similar image in each respective first set of similar images:attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold;determining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image; andclassifying only the similar images that are determined to belong the first category as belonging to the first category of images;wherein the first category of images are images that have been classified as spoof images. 12. The system of claim 11, wherein: each seed image is an image for which a classification score generated based on the signal data meets a first classification score threshold that corresponds to a first likelihood that an image belongs to the first category; anddetermining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image comprises: generating a classification score based on the signal data attributed to the similar image; anddetermining that the classification score meets at least a second classification score threshold that corresponds to a second likelihood that an image belongs to the first category, and wherein the second likelihood is less than the first likelihood. 13. The system of claim 11, the operations further comprising: for each first seed image: determining a respective first set of similar seed images from the other first seed images, each similar seed image in the respective first set of similar seed images having a visual similarity score that is a measure of visual similarity of the similar seed image to the first seed image based on the image content of similar image and the first seed image, and that satisfies a first seed image similarity threshold; andwherein, for each similar image in each respective first set of similar images, attributing to the similar image the signal data of each first seed image comprises: attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold, and the signal data attributed to each first seed image from the similar seed images. 14. The system of claim 11, the operations further comprising: accessing data identifying a set of second seed images, each second seed image being classified as not belonging to the first category of images based on signal data of the second seed image that are independent of image content of the second seed image, and wherein the set of second seed images is a proper subset of images in an image corpus;for each second seed image: determining a respective second set of similar images from images in the image corpus, each similar image in the respective second set of images having a visual similarity score that is a measure of visual similarity of the similar image to the second seed image based on the image content of the similar image and the second seed image, and that satisfies a first seed image similarity threshold; andfor each similar image in each respective second set of similar images: attributing to the similar image the signal data of each second seed image for which the similar image has a respective visual similarity score satisfying the second seed image similarity threshold; andwherein determining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image comprises, for each similar image belonging a respective first set and a respective second set, determining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image from the first seed images and the second seed images. 15. The system of claim 11, wherein the signal data of the first seed images comprise hover metric data indicative of one or more hover metrics for the first seed image. 16. The system of claim 11, wherein attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold comprises adjusting signal data of the first seed image attributed to the similar image by a value that is proportional to the visual similarity score of the similar image. 17. A non-transitory memory storage apparatus storing instructions executable by a data processing apparatus and that upon such execution cause the data processing apparatus to perform operations comprising: accessing data identifying a set of first seed images, each first seed image being classified as belonging to a first category of images based on signal data of the first seed image that are independent of image content of the first seed image, and wherein the set of first seed images is a proper subset of images in an image corpus;for each first seed image:determining a respective first set of similar images from images in the image corpus, each similar image in the respective first set of images having a visual similarity score that is a measure of visual similarity of the similar image to the first seed image based on the image content of the similar image and the first seed image, and that satisfies a first seed image similarity threshold; andfor each similar image in each respective first set of similar images:attributing to the similar image the signal data of each first seed image for which the similar image has a respective visual similarity score satisfying the first seed image similarity threshold;determining whether the similar image belongs to the first category of images based on the image signal data attributed to the similar image; andclassifying only the similar images that are determined to belong the first category as belonging to the first category of images;wherein the first category of images are images that have been classified as spoof images.
Barber Ronald J. (San Jose CA) Beitel Bradley J. (Woodside CA) Equitz William R. (Palo Alto CA) Niblack Carlton W. (San Jose CA) Petkovic Dragutin (Saratoga CA) Work Thomas R. (San Francisco CA) Yank, Image query system and method.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.