Fashion-Product-Classifier

Overview

This project builds and deploys a deep learning model to predict multiple attributes of fashion products, including:

• Article Type (e.g., T-shirt, Jeans, Shoes)
• Base Colour (e.g., Red, Blue, Black)
• Season (e.g., Summer, Winter)
• Gender (e.g., Men, Women, Unisex)

The model is trained using MobileNetV2 as a feature extractor and fine-tuned for multi-task classification. It is deployed as a REST API using Flask or FastAPI.

Dataset

• Source: Fashion Product Images Dataset - Kaggle
• Trained Model: https://drive.google.com/file/d/14k1ifBlwHVOUI42voODcio30mnlRefS0/view?usp=sharing (Pickel File uploaded here)
• Data: Contains images of fashion products and a CSV file (styles.csv) with labels for product type, color, season, and gender.

Features

• Data Preprocessing: Exploratory Data Analysis (EDA), Filtering Rows with Missing Images, Encoding Categorical Labels,Splitting the Dataset and image augmentation.
• Multi-Task Deep Learning Model: Uses MobileNetV2 as a base model.
• Model Training & Optimization: Trained with categorical cross-entropy loss and Adam optimizer.
• Deployment: The trained model is saved (.h5 format) and served via an API.
• Dockerization: Containerized using Docker for easy deployment.

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/fashion-product-prediction.git
   cd fashion-product-prediction

2. Install dependencies:

   pip install -r requirements.txt

3. Download and extract the dataset from Kaggle.

4. Train the model:

   fashion_product_attribute_prediction.ipynb

5. Run the API (FastAPI or Flask):

     app(flask).py (or) app(streamlit).py

Model Training

• The MobileNetV2 architecture is used with additional dense layers for multi-task classification.
• Data augmentation is applied using ImageDataGenerator.
• Model is trained for 10 epochs using batch size 32.

API Usage

Once deployed, you can send an image to the API endpoint /predict:

Example Request (Using cURL)

curl -X POST "http://localhost:5000/predict" \
     -H "Content-Type: multipart/form-data" \
     -F "image=@path/to/image.jpg"

Example Response

{
  "articleType": "Jeans",
  "baseColour": "Blue",
  "season": "Winter",
  "gender": "Men"
}

Deployment with Docker

1. Build the Docker image:

   docker build -t fashion-predictor .

2. Run the container:

   docker run -p 5000:5000 fashion-predictor

Results

• The model achieves high accuracy across all categories.
• It can predict multiple attributes from a single image input.

Contributors

• M Karthikeya Reddy (GitHub)

License

This project is licensed under the MIT License. Feel free to contribute and improve it!

---

🔗 GitHub Repo: [(https://github.com/Karthikeyareddy183/Fashion-Product-Classifier)]