rzamarefat/SelfSupervised_Visual_Search_ECommerce

A fullstack vision-based web application for recommending similar products

Self-Supervised Visual Search Ecommerce

Description

A fullstack vision-based web application for recommending similar products

Demo

Technologies

• Self-Supervised approaches: SimCLR, BYOL
• Deep Learning: Pytorch
• Similarity search: FAISS
• Backend: FastAPI
• Frontend: React JS and Redux
• Mark-up and style: HTML/CSS/Bootstrap

Overall procedure

Instruction to set-up

• In the src directory you can find the related setup instruction for back and frontend seperately.
• It requires common steps taken in any React and FastAPI app. So don't worry.

Dataset:

• Fashion Product Images Dataset

Pre-Trained weights:

• In the src/back directory you can find a file named 'epoch=19-step=2960.ckpt'

Refs:

• SimCLR paper
• BYOL paper
• For SimCLR implementation, I have lost the link and now I can't find where I previously saw it.
• SimCLR - AI Summer
• BYOL - AI Summer