Car Damage Detection API

A FastAPI application for car damage detection using a YOLO-based model.

Features

• Upload an image to detect car damage
• Returns a processed image with bounding boxes around detected damage
• Provides detailed information about detected objects including class, confidence, and coordinates
• Simple health check endpoint to verify API is operational

Installation

1. Clone this repository:

   git clone https://github.com/amangupta20/carddapi.git
   cd carddapi
   

2. Create a virtual environment and activate it:

   python -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

3. Install the dependencies:

   pip install -r requirements.txt
   

Usage

1. Start the FastAPI server:

   uvicorn app:app --host 0.0.0.0 --port 8000 --reload
   

2. The API will be available at http://localhost:8000

3. Access the interactive API documentation at http://localhost:8000/docs

API Endpoints

GET /

• Returns a welcome message to confirm the API is running

GET /health

• Checks if the API is healthy and if the model is loaded properly

POST /detect

• Endpoint for car damage detection
• Parameters:
  • file: Image file to be analyzed (multipart/form-data)
  • confidence: Confidence threshold (optional, default: 0.25)
• Returns:
  • processed_image: Base64 encoded image with bounding boxes
  • inference_time: Time taken for inference in seconds
  • total_objects: Total number of objects detected
  • detections: List of detection results with class, confidence, and coordinates
  • class_counts: Count of each class detected

Example

Using curl:

curl -X POST "http://localhost:8000/detect" \
     -H "accept: application/json" \
     -H "Content-Type: multipart/form-data" \
     -F "file=@path/to/your/image.jpg" \
     -F "confidence=0.25"

Using Python requests:

import requests

url = "http://localhost:8000/detect"
files = {"file": open("path/to/your/image.jpg", "rb")}
data = {"confidence": 0.25}

response = requests.post(url, files=files, data=data)
result = response.json()

# The result contains:
# - processed_image: Base64 encoded image with bounding boxes
# - inference_time: Time taken for inference
# - total_objects: Number of objects detected
# - detections: List of detection results
# - class_counts: Count of each class detected

Docker Support (Optional)

To run the API in a Docker container:

1. Build the Docker image:

   docker build -t car-damage-detection .
   

2. Run the container:

   docker run -p 8000:8000 car-damage-detection