Clothes_Search_Engine

1. Project Overview

The Clothes Search Engine project is a web application designed to help people find clothes that are similar to the input clothes.

2. Project Components

2.1. DeepFashion Dataset

• Link dataset: https://mmlab.ie.cuhk.edu.hk/projects/DeepFashion.html.

2.2. Detection Model

• Use a clothes detection model from the existing project on GitHub.
  • Link: https://github.com/TanThinNguyen/fashion-visual-search.

2.3. Project Pipeline

• Using Detection Model to get the clothes in the image and then crop the clothes that have the highest detection score.
• Using ResNet50 neural network to extract features from the image and save to a .npy file.
• Using FAISS(Facebook AI Similarity Search) to save all .npy file to search.
• Building User Interface.

3. Getting started

3.1. Requirements

Main libraries:

• Python version: 3.8.18.
• Tensorflow Object Detection API: Install it in notebook.ipynb.
• Numpy version: 1.24.4.
• Flask: 3.0.3.
• faiss-gpu version: 1.7.2.
• faiss-cpu version: 1.8.0.
• Tensorflow Version: 2.13.1.
• Pillow Version: 10.4.0.

3.2. Getting Dataset

• Access to the Google Drive link of the DeepFashion Dataset then open the folder "Category and Attribute Prediction Benchmark". After that, download file img.zip, extract and save it into folder static in the project folder.

3.3. Running The Project

1. Readding notebook.ipynb to get the detection model and faiss_index (download and save it into the project folder).
2. Run the web app:

python main.py 

4. Preview

• Front Page: $\newline$

• Result Page: $\newline$