PicSort - Face-Based Image Sorter API

PicSort is a FastAPI-based mini backend that takes a zipped folder of images, detects and analyzes faces using DeepFace, and returns a zipped folder of images grouped into folders by unique faces. It also detects and separates images with:

• No faces (in a no_faces folder)
• Invalid/corrupt files (in an invalid folder)

---

Features

• Upload a zipped folder of images
• Automatically detect and group similar faces into individual folders
• Separate:
  • Images with no face
  • Corrupted or unreadable files
• Download a zipped, sorted version of your original dataset
• In-memory face matching using DeepFace + Facenet (no database required)
• CORS-enabled for frontend integration

---

Tech Stack

• Python 3.8+
• FastAPI
• DeepFace (with Facenet and VGG model)
• OpenCV, shutil
• Uvicorn (ASGI server)

---

Getting Started

1. Clone the Repository

git clone https://github.com/5-Abhi-5/PicSort.git
cd picsort

2. Create a virtual environment

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

3. Install required packages

pip install -r requirements.txt

4. Run the App

uvicorn main:app --reload

---

Endpoint

🔹 POST /uploadfile — Upload a Zipped Folder of Images

• Description: Upload a .zip file containing folder with images. The server processes and groups them by face similarity.
• Request Body: multipart/form-data
  • Field: file (type: File)
• Response: 200 OK with downloadable zipped file.

Curl Example (localhost)

curl -X POST "http://127.0.0.1:8000/uploadfile" \
  -H "accept: application/json" \
  -H "Content-Type: multipart/form-data" \
  -F "file=@/images_path.zip;type=application/x-zip-compressed"