Cloth Similarity Recommendation

APP Link : https://cloth-recommendation.streamlit.app/

Overview

This project provides a clothing similarity recommendation system using a deep learning feature extractor and a FastAPI backend. A pre-trained MobileNet model is fine-tuned on a custom dataset to extract image features, which are stored in a SQLite database. Users can upload images to find visually similar clothing items.

Prerequisites

• Python 3.8 or higher
• pip
• Virtual environment tool (optional but recommended)

Setup

1. Clone the repository:

   git clone <repo_url>
   cd "Cloth Similarity"

2. Create and activate a virtual environment:

   python -m venv cloth_env
   # Windows
   cloth_env\Scripts\activate.bat
   # macOS/Linux
   source cloth_env/bin/activate

3. Install required packages:

   pip install -r requirements.txt

Training the Feature Extractor

1. Launch Jupyter Notebook:

   jupyter notebook exp.ipynb

2. In exp.ipynb:
   • Run cells to load and preprocess data from dataset/train and dataset/test.
   • Perform data augmentation with ImageDataGenerator.
   • Load MobileNet base model (weights from ImageNet) and build a custom classifier on top.
   • Train the model for 10 epochs.
   • Fine-tune by unfreezing the base model and re-running training for another 10 epochs.
   • Save the feature extractor:

     feature_extractor.save('feature_extractor.keras')

API Backend

The FastAPI app (api.py) provides two endpoints:

• POST /reset_and_update_db

  • Deletes existing textile.db if present.
  • Crawls product images from the configured base URL.
  • Extracts features using the saved feature extractor and populates the SQLite database.
  • Returns a success message.

• POST /upload

  • Accepts an image file upload.
  • Preprocesses and extracts features.
  • Queries the database for the top 5 visually similar items using cosine similarity.
  • Returns the list of image URLs with similarity scores.

Running the API

uvicorn api:app --reload

The API will be available at http://127.0.0.1:8000.

Usage

1. Reset and populate the database:

   curl -X POST http://127.0.0.1:8000/reset_and_update_db

2. Upload an image to get recommendations:

   curl -X POST -F "file=@path/to/image.jpg" http://127.0.0.1:8000/upload

Dataset Structure

dataset/
  train/
    pants/
    shirt/
    shorts/
    t-shirt/
  test/
    pants/
    shirt/
    shorts/
    t-shirt/

License

This project is licensed under the MIT License.