Detecting similarity between images

1. A computer-implemented method comprising:

receiving a request to display an interface including an item that has an associated image;

requesting for, and receiving, a determination one or more similarity scores between the item and items of a same or similar category as the item, the one or more similarity scores based upon one or more Gram matrix-based style values and one or more vector distance calculation-based content values as determined from one or more outputs of layers of a convolutional neural network; and

displaying one or more images of items that are similar to the image based on the one or more similarity scores.

2. The computer-implemented method of claim 1 , wherein the convolutional neural network is a pre-trained image feature extraction model.

3. The computer-implemented method of claim 1 , wherein which layers of the convolutional neural network to use for the Gram matrix-based and vector distance calculation-based content values are configurable.

4. A computer-implemented method comprising:

receiving a request to determine similarity between a first image and at least one other image;

determining similarity between the first image and at least one other image based upon one or more Gram matrix-based style values and one or more vector distance calculation-based content values as determined from one or more outputs of layers of a convolutional neural network; and

providing an indication of the similarity between the first image and the at least one other image.

5. The computer-implemented method of claim 4 , wherein the convolutional neural network is a pre-trained image feature extraction model.

6. The computer-implemented method of claim 5 , wherein the vector distance calculation is one of L1 or L2 distance.

7. The computer-implemented method of claim 4 , wherein which layers of the convolutional neural network to use for the Gram matrix-based and vector distance calculation-based content values are configurable.

8. The computer-implemented method of claim 4 , wherein the indication is a similarity value generated by determining a difference of blended one Gram matrix-based style values and one or more vector distance calculation-based content values of the first image and blended one Gram matrix-based style values and one or more vector distance calculation-based content values of the at least one other image.

9. The computer-implemented method of claim 4 , wherein the similarity between the first image and at least one other image is determined by an indexed database query to find images having similar Gram-based style values and vector distance calculation-based content values as the first image.

10. The computer-implemented method of claim 9 , wherein the indication is one or more locations of images that have similar Gram-based style values and vector distance calculation-based content values.

11. The computer-implemented method of claim 4 , wherein the similarity between the first image and at least one other image is determined by a nearest neighbor search against an index to find images having similar Gram-based style values and vector distance calculation-based content values as the first image.

12. The computer-implemented method of claim 4 , wherein the indication is one or more locations of images that have similar Gram-based style values and L2-based content values.

13. The computer-implemented method of claim 4 , wherein the convolutional neural network has been configured using a configuration having on or more of: content extraction layers identification and respective layer weightings, style extraction layers identification and respective layer weightings, a ratio of content versus style influence, and image storage information.

14. The computer-implemented method of claim 4 , further comprising:

receiving an input to change a configuration of which layers of the convolutional neural network are to be used to determine similarity between the first image and at least one other image.

15. A system comprising:

a web application implemented by a first one or more electronic devices; and

a neural style inference service implemented by a second one or more electronic devices, the neural style inference service including instructions that upon execution cause the neural style inference service to: receive a request from the web application to determine similarity between a first image and at least one other image;

determine similarity between the first image and at least one other image based upon one or more Gram matrix-based style values and one or more vector distance calculation-based content values as determined from one or more outputs of layers of a convolutional neural network; and

provide an indication of the similarity between the first image and the at least one other image.

16. The system of claim 15 , wherein the convolutional neural network is a pre-trained image feature extraction.

17. The system of claim 15 , wherein the vector distance calculation is one of L1 or L2 distance.

18. The system of claim 15 , wherein which layers of the convolutional neural network to use for the Gram matrix-based and vector distance calculation-based content values are configurable.

19. The system of claim 15 , wherein the similarity between the first image and at least one other image is determined by a nearest neighbor search against an index to find images having similar Gram-based style values and vector distance calculation-based content values as the first image.

20. The system of claim 15 , wherein the convolutional neural network has been configured using a configuration having on or more of: content extraction layers identification and respective layer weightings, style extraction layers identification and respective layer weightings, a ratio of content versus style influence, and image storage information.