Image Search ML Service

A machine learning service for image-based auto parts product search. Upload a photo of a car part and get matching products from the catalog.

Features

• Image Search by Upload: Send a product image and receive ranked matching results
• YOLOv8 Part Detection: Automatically detects auto part categories (brake pads, filters, etc.)
• OCR Text Extraction: Reads brand names and part numbers from images using PaddleOCR
• CLIP Image Embeddings: Semantic similarity search via vector embeddings
• FAISS Vector Index: Fast approximate nearest neighbor search for large catalogs
• Dynamic Result Merging: Combines catalog search + vector search with confidence-weighted scoring

Tech Stack

• FastAPI - Web framework
• YOLOv8 (Ultralytics) - Object detection for part types
• PaddleOCR - Optical character recognition
• OpenAI CLIP - Image/text embeddings
• BGE-M3 - Text embeddings for semantic search
• FAISS - Vector similarity search
• PyTorch - ML backend

Quick Start

Local Development

1. Clone and setup environment:

   git clone https://github.com/services-management/image-search.git
   cd image-search
   python -m venv venv
   source venv/bin/activate  # Windows: venv\Scripts\activate
   pip install -r requirements.txt

2. Configure environment:

   cp .env.example .env
   # Edit .env with your settings

3. Run the service:

   uvicorn app.main:app --reload --port 8001

Docker

docker-compose up -d

The service will be available at http://localhost:8001.

API Endpoints

Endpoint	Method	Description
/	GET	Service info
/health	GET	Health check
/api/v1/search-by-image	POST	Search products by image upload
/api/v1/index-product	POST	Add a product image to the search index
/api/v1/rebuild-index	POST	Rebuild the FAISS index from catalog
/api/v1/index/stats	GET	FAISS index statistics
/api/v1/brands	GET	List known auto part brands
/api/v1/categories	GET	List supported part categories

Example Usage

curl -X POST http://localhost:8001/api/v1/search-by-image \
  -F "file=@brake_pad.jpg" \
  -F "top_k=10"

Testing

pytest

Project Structure

app/                # FastAPI application
pipeline/           # ML pipelines (YOLO, OCR, CLIP, embeddings)
search/             # FAISS index, catalog client, result merger
tests/              # Test suite
scripts/            # Training and utility scripts

Environment Variables

See .env.example for all available configuration options.

License

Private - services-management