Method of identifying similars in clothing images

1. A system for identifying similars for a selected image amongst a set of clothing images in an electronic catalog, comprising:

a processor; and

a memory containing instructions, when executed by the processor, configure the system to:

apply an object detection model to extract a clothing section within said clothing images to create a preprocessed image;

apply a first machine learning model on said preprocessed image(s) to convert colors and textures of said preprocessed image into a first set of vector representations and thereafter store said first set of vector representations in a vector database;

apply a second machine learning model on said preprocessed image(s) to convert shapes of said preprocessed image into a second set of vector representations and thereafter store said second set of vector representations in said vector database;

map and identify a set of n nearest vectors of said first set of vectors with the vector of said selected image from said vector database;

map and identify a set of m nearest vectors of said second set of vectors with the vector of said selected image from said vector database;

identify a set of n images corresponding to said set of n nearest vectors;

identify a set of m images corresponding to said set of m nearest vectors;

apply an attribute extraction model to identify clothing attributes of said selected image;

apply an attribute extraction model to identify clothing attributes of each image in each set of said identified sets of n and m images;

match attributes between said identified clothing attributes of said selected image and said identified clothing attributes of each image in each set of said identified sets of n and m images;

sort and rank, based on said match, each image of said identified set of n images and each image of said identified set of m images; and

display on a user interface said sorted and ranked sets of n and m images.

2. The system of claim 1 , wherein said clothing attributes comprises one or more of design of collar, shape of collar, shape of sleeve, and text and numbering content of said preprocessed image.

3. The system of claim 1 , wherein said first machine learning model is a Bootstrap Your Own Latent (BYOL) model.

4. The system of claim 3 , wherein said BYOL model is trained to create different views of the same image utilizing a random cropping augmentation technique.

5. The system of claim 3 , wherein said BYOL model is trained to create different views of the same image utilizing a left-right flip augmentation technique.

6. The system of claim 3 , wherein said BYOL model is trained to create different views of the same image utilizing one or more of a color jittering and color dropping augmentation technique.

7. The system of claim 1 , wherein said second machine learning model is a non-parametric instance-level discrimination model.

8. The method of claim 1 , wherein the system is further configured to:

select images common to said sorted and ranked sets of n and m images; and

display said selected common images on said user interface.

9. A method for identifying similars for a selected clothing image amongst a set of clothing images in an electronic catalog, comprising:

applying a object detection model to extract a clothing section within said clothing images to create a preprocessed image;

applying a first machine learning model on said preprocessed image(s) to convert colors and textures of said preprocessed image into a first set of vector representations and thereafter store said first set of vector representations in a vector database;

applying a second machine learning model on said preprocessed image(s) to convert shapes of said preprocessed image into a second set of vector representations and thereafter store said second set of vector representations in said vector database;

applying a non-parametric instance-level discrimination model to generate a second set of vector representation of clothing images in the electronic catalog to determine a second set of similar images;

mapping and identifying a set of n nearest vectors of said first set of vectors with the vector of said selected image from said vector database;

mapping and identifying a set of m nearest vectors of said second set of vectors with the vector of said selected image from said vector database;

identifying a set of n images corresponding to said set of n nearest vectors;

identifying a set of m images corresponding to said set of m nearest vectors;

applying an attribute extraction model to identify clothing attributes of said selected image;

applying an attribute extraction model to identify clothing attributes of each image in each set of said identified sets of n and m images;

matching attributes between said identified clothing attributes of said selected image and said identified clothing attributes of each image in each set of said identified sets of n and m images;

sorting and ranking, based on said match, each image of said identified set of n images and each image of said identified set of m images; and

displaying on a user interface said sorted and ranked sets of n and m images.

10. The method of claim 9 , wherein said clothing attributes comprises one or more of design of collar, shape of collar, shape of sleeve, and text and numbering content of said preprocessed image.

11. The method of claim 9 , wherein said first machine learning model is a Bootstrap Your Own Latent (BYOL) model.

12. The method of claim 11 , wherein said BYOL model is trained to create different views of the same image utilizing a random cropping augmentation technique.

13. The method of claim 11 , wherein said BYOL model is trained to create different views of the same image utilizing a left-right flip augmentation technique.

14. The method of claim 11 , wherein said BYOL model is trained to create different views of the same image utilizing one or more of a color jittering and color dropping augmentation technique.

15. The method of claim 9 , wherein said second machine learning model is a non-parametric instance-level discrimination model.