System and method for facilitating graphic-recognition training of a recognition model
Methods and computer readable media for facilitating training of a recognition model. An embodiment includes generating media items based on information associated with a representation of a graphic, the information including content other than the graphic, content based on at least one transformation parameter set, and content comprising the graphic integrated with the other content, then using a recognition model to process the media items to generate predictions related to recognition of the graphic for the media items, the generated predictions including an indication of a predicted location of the graphic in a first media item. The process also includes presenting an indication of the predicted location on an area of the first media item via a user interface to a user, then obtaining a reference feedback set that includes reference indications related to recognition of the graphic for the media items and including user feedback concerning the indication of the predicted location of the graphic, and then updating the recognition model based on the reference feedback.
1. A method for facilitating training of a recognition model comprising:
generating media items based on information associated with a representation of a graphic, the information comprising content other than the graphic, content based on at least one transformation parameter set, and content comprising the graphic integrated with other content;
generating, by using a recognition model to process the media items, predictions related to recognition of the graphic for the media items, the generated predictions comprising an indication of a predicted location of the graphic in a first media item;
presenting, via a user interface to a user, an indication of the predicted location on an area of the first media item;
obtaining a reference feedback set, the reference feedback set comprising reference indications related to recognition of the graphic for the media items and including user feedback via the user interface concerning the indication of the predicted location of the graphic on an area of the first media item; and
updating the recognition model based on the reference feedback set including the user feedback concerning the indication of the predicted location of the graphic.
2. The method of claim 1 , wherein generating the media items comprises using a plurality of transformation parameter sets, wherein each of the transformation parameter sets comprise different parameters from each other,
and wherein generating the media items comprises generating at least some media items by applying a different one of the transformation parameter sets to the representation of the graphic such that each media item of the at least some media items has a different transformed representation of the graphic from one another.
3. The method of claim 2 , wherein obtaining the transformation parameter sets comprises randomly generating at least some of the transformation parameter sets.
4. The method of claim 2 , wherein the transformation parameter sets comprise occlusion parameters, and
wherein generating the at least some media items comprises generating one or more of the at least some media items based on the occlusion parameters such that each media item has a transformed representation of the graphic in which at least of a portion of the representation of the graphic is at least one of missing or hidden.
5. The method of claim 2 , wherein the transformation parameter sets comprise at least one of blurring effect parameters, camera effect parameters, motion effect parameters, shadow effect parameters, pattern effect parameters, or texture effect parameters, sharpening parameters, softening parameters, brightness parameters, contrast parameters, or recoloring parameters.
6. The method of claim 2 , wherein the transformation parameter sets comprise compression parameters.
7. The method of claim 1 , wherein the recognition model comprises a neural network, and wherein each of the media items is an image.
8. The method of claim 1 , wherein generating the media items further comprises generating the first media item such that at least some content other than the graphic appears opaquely over at least a portion of the graphic on the first media item, and
wherein generating the predictions comprises generating the indication of the predicted location based on the first media item in which the at least some content opaquely appears over at least a portion of the graphic.
9. The method of claim 1 , wherein obtaining the reference indications comprises:
obtaining a reference indication of an object to be recognized via the user interface by the user of the indication of the predicted location, and
wherein updating the recognition model comprises updating the recognition model based on the reference indication.
10. The method of claim 1 , further comprising:
generating representations of the graphic such that each of the representations of the graphic has a different size from one another,
wherein generating the media items comprises generating at least some media items such that each media item of the at least some media items comprises a different representation of the graphic.
11. The method of claim 1 , wherein the recognition model:
determines similarities or differences between the generated predictions and their corresponding reference indications, and
updates neural network aspects of the recognition model based on the determined similarities or differences.
12. A non-transitory, computer-readable media storing instructions for facilitating training of a recognition model that, when executed by a one or more processors, cause operations comprising:
generating media items based on information associated with a representation of a graphic, the information comprising content other than the graphic, content based on at least one transformation parameter set, and content comprising the graphic integrated with other content;
generating, by using a recognition model to process the media items, predictions related to recognition of the graphic for the media items, the generated predictions comprising an indication of a predicted location of the graphic in a first media item;
presenting, via a user interface to a user, an indication of the predicted location on an area of the first media item;
obtaining a reference feedback set, the reference feedback set comprising reference indications related to recognition of the graphic for the media items and including user feedback via the user interface concerning the indication of the predicted location of the graphic on an area of the first media item; and
updating the recognition model based on the reference feedback set including the user feedback concerning the indication of the predicted location of the graphic.
13. The non-transitory, computer-readable media of claim 12 , wherein generating the media items comprises using a plurality of transformation parameter sets, wherein each of the transformation parameter sets comprise different parameters from each other,
and wherein generating the media items comprises generating at least some media items by applying a different one of the transformation parameter sets to the representation of the graphic such that each media item of the at least some media items has a different transformed representation of the graphic from one another.
14. The non-transitory, computer-readable media of claim 13 , wherein obtaining the transformation parameter sets comprises randomly generating at least some of the transformation parameter sets.
15. The non-transitory, computer-readable media of claim 13 , wherein the transformation parameter sets comprise occlusion parameters, and
wherein generating the at least some media items comprises generating one or more of the at least some media items based on the occlusion parameters such that each media item has a transformed representation of the graphic in which at least of a portion of the representation of the graphic is at least one of missing or hidden.
16. The non-transitory, computer-readable media of claim 13 , wherein the transformation parameter sets comprise at least one of blurring effect parameters, camera effect parameters, motion effect parameters, shadow effect parameters, pattern effect parameters, or texture effect parameters, sharpening parameters, softening parameters, brightness parameters, contrast parameters, or recoloring parameters.
17. The non-transitory, computer-readable media of claim 13 , wherein the transformation parameter sets comprise compression parameters.
18. The non-transitory, computer-readable media of claim 12 , wherein generating the media items further comprises generating the first media item such that at least some content other than the graphic appears opaquely over at least a portion of the graphic on the first media item, and
wherein generating the predictions comprises generating the indication of the predicted location based on the first media item in which the at least some content opaquely appears over at least a portion of the graphic.
19. The non-transitory, computer-readable media of claim 12 , wherein obtaining the reference indications comprises:
obtaining a reference indication of an object to be recognized via the user interface by the user of the indication of the predicted location, and
wherein updating the recognition model comprises updating the recognition model based on the reference indication.
20. The non-transitory, computer-readable media of claim 12 , further comprising:
generating representations of the graphic such that each of the representations of the graphic has a different size from one another,
wherein generating the media items comprises generating at least some media items such that each media item of the at least some media items comprises a different representation of the graphic.