Device and method for item recommendation based on visual elements

1. A method for item recommendation based on visual elements, the method comprising:

determining, by one or more processors, visual elements from an item image of an item;

analyzing, by the one or more processors, the determined visual elements so as to determine feature elements from the determined visual elements, in response to the item being collocated with one or more other items, wherein the feature elements comprise all or part of the determined visual elements;

generating, by the one or more processors, an element descriptor for the item based on at least a part of the visual elements; and

calculating, by the one or more processors, a compatibility value between the element descriptor and one or more other element descriptors for the one or more other items,

wherein the step of generating the element descriptor for the item based on at least a part of the visual elements comprises:

generating, by the one or more processors, the element descriptor based on the feature elements.

2. The method of claim 1 , further comprising:

comparing, by the one or more processors, the compatibility value with a compatibility threshold.

3. The method of claim 2 , further comprising:

recommending, by the one or more processors, the one or more other items in response to the compatibility value being greater than or equal to the compatibility threshold.

4. The method of claim 1 , wherein the feature elements are determined by two Convolutional long short term memory (ConvLSTM) Networks with attention mechanism.

5. The method of claim 1 , wherein before the step of determining visual elements from an item image of an item, the method further comprises steps of:

acquiring, by the one or more processors, a raw image comprising the item image;

performing, by the one or more processors, human parsing on the raw image to generate a parsed image; and

generating, by the one or more processors, the item image based on the raw image and the parsed image.

6. The method of claim 5 , wherein the step of generating the item image based on the raw image and the parsed image comprises steps of:

cropping, by the one or more processors, the parsed image to generate a cropped parsed image comprising a cropped item region and a cropped non-item region;

filtering out, by the one or more processors, the cropped non-item region from the cropped parsed image, to generate a human parsing mask; and

overlaying, by the one or more processors, the human parsing mask with the raw image to generate the item image.

7. The method of claim 5 , wherein the step of performing human parsing comprises a step of:

executing, by the one or more processors, a human parsing algorithm that is trained on datasets,

wherein the human parsing algorithm is visual-keypoint and/or human-keypoint aided human parsing algorithm.

8. The method of claim 1 , wherein before the step of generating the element descriptor for the item, the method further comprising:

pruning, by the one or more processors, from the visual elements, a visual element that does not satisfy any of the following conditions:

a) overlapping with Scale-invariant Feature Transform (SIFT)/Speed Up Robust Feature (SURF)/Maximally Stable Extremal Region (MSER) keypoints; and

b) overlaying with visual landmarks.

9. A device for item recommendation based on visual elements, the device comprises:

a processor;

a memory storing instructions which, when executed by the processor, cause the processor to:

determine visual elements from an item image of an item;

analyze the determined visual elements so as to determine feature elements from the determined visual elements, in response to the item being collocated with one or more other items, wherein the feature elements comprise all or part of the determined visual elements;

generate an element descriptor for the item based on at least a part of the visual elements; and

calculate a compatibility value between the element descriptor and one or more other element descriptors for the one or more other items,

wherein the instructions which, when executed by the processor, further cause the processor to:

generate the element descriptor based on the feature elements.

10. The device of claim 9 , wherein the instructions which, when executed by the processor, further cause the processor to:

compare the compatibility value with a compatibility threshold.

11. The device of claim 10 , wherein the instructions which, when executed by the processor, further cause the processor to:

recommend the one or more other items in response to the compatibility value being greater than or equal to the compatibility threshold.

12. The device of claim 9 , wherein the feature elements are determined by two Convolutional long short term memory (ConvLSTM) Networks with attention mechanism.

13. The device of claim 9 , wherein the instructions which, when executed by the processor, further cause the processor to:

acquire a raw image comprising the item image;

perform human parsing on the raw image to generate a parsed image; and

generate the item image based on the raw image and the parsed image.

14. The device of claim 13 , wherein the instructions which, when executed by the processor, further cause the processor to:

crop the parsed image to generate a cropped parsed image comprising a cropped item region and a cropped non-item region;

filter out the cropped non-item region from the cropped parsed image, to generate a human parsing mask; and

overlay the human parsing mask with the raw image to generate the item image.

15. The device of claim 9 , wherein the instructions which, when executed by the processor, further cause the processor to:

prune, from the visual elements, a visual element that does not satisfy any of the following conditions:

a) overlapping with Scale-invariant Feature Transform (SIFT)/Speed Up Robust Feature (SURF)/Maximally Stable Extremal Region (MSER) keypoints; and

b) overlaying with visual landmarks.

16. A non-transitory computer readable medium storing computer executable instructions which, when executed by a processor of a computing device, causes the processor to:

determine visual elements from an item image of an item;

analyze the determined visual elements so as to determine feature elements from the determined visual elements, in response to the item being collocated with one or more other items, wherein the feature elements comprise all or part of the determined visual elements;

generate an element descriptor for the item based on at least a part of the visual elements; and

calculate a compatibility value between the element descriptor and one or more other element descriptors for the one or more other items,

wherein the instructions which, when executed by the processor, further cause the processor to:

generate the element descriptor based on the feature elements.