Deep Image Search API

This FastAPI-based API provides a simple and efficient way to search for similar images using the DeepSearchLite library. The API receives an image and returns a list of similar images based on a trained feature extractor model. This API serves as the backend for the portfolio web app for the RevImageSearch project, which can be accessed at My Portfolio.

Setup

1. Install required packages:

   pip install fastapi uvicorn Pillow
   

2. Download the trained feature extractor model and place it in a models directory.

3. Configure the MODEL_PATH, IMAGES_DIR, METADATA_DIR, and FEATURE_EXTRACTOR_NAME variables according to your setup.

4. Run the API:

   uvicorn app.main:app --host 0.0.0.0 --port 8000
   

Usage

Send a POST request to the /search-similar-images endpoint with an image file and an optional number of similar images to return. Example:

curl -X POST "http://localhost:8000/search-similar-images" -H "accept: application/json" -H "Content-Type: multipart/form-data" -F "image=@path_to_image.jpg;type=image/jpeg" -F "number_of_images=3"

The API will return a JSON response containing a list of similar images in base64 format:

{
  "similar_images": [
    "data:image/jpeg;base64,...",
    "data:image/jpeg;base64,...",
    "data:image/jpeg;base64,..."
  ]
}

You can easily visualize the base64 images in web applications or convert them back to image files using various libraries.

---