README.md

metadata

title: Deep Detect Api
emoji: 🦀
colorFrom: red
colorTo: blue
sdk: docker
app_port: 7860

Image_Detector - FastAPI Backend

This directory houses the PyTorch computer vision backend service for Deep-Detect. It exposes a robust FastAPI endpoint that accepts image payloads and predicts whether the image is Real or Deep-Fake (AI-generated) using a custom Convolutional Neural Network (CNN).

---

Directory Structure

Image_Detector/
├── app.py              # FastAPI application - server entry point, CORS, and routes
├── inference.py        # Model loader, transform pipeline, and prediction execution
├── model.py            # Custom CNN architecture structure (PyTorch definition)
├── predict.py          # Tkinter desktop desktop utility for local file scans
├── requirements.txt    # Python module dependencies
├── models/             # Target folder for model weights (gitignored)
│   └── custom_cnn_standalone.pt
└── notebooks/          # Research and model training steps
    ├── Preprocessing.ipynb
    ├── Model_training.ipynb
    ├── Model_evaluation.ipynb
    └── Pretrained_Models.ipynb

---

Technical Specifications

• Model Framework: PyTorch (compiled and exported as TorchScript .pt file).
• Classification type: Binary (Class 0: AI/Deep-Fake, Class 1: Real).
• Input Dimension: 224 x 224 pixels, 3 channels (RGB).
• Inference logic: Logit output -> Sigmoid function -> Probability.
• Decision boundary: Probability threshold of 0.5.
  • Probability > 0.5 -> Real (Class 1)
  • Probability <= 0.5 -> Deep-Fake/AI (Class 0)
• Confidence Computation:
  • For Real: probability * 100
  • For AI: (1.0 - probability) * 100

---

Setup and Installation

Prerequisites

• Python 3.10 or higher.
• pip package manager.

Steps

1. Activate Virtual Environment:

   cd Image_Detector
   python -m venv venv
   
   # Windows (PowerShell)
   .\venv\Scripts\activate
   
   # Linux / macOS
   source venv/bin/activate
   

2. Install Dependencies:

   pip install -r requirements.txt
   

3. Provide Weight File: Download custom_cnn_standalone.pt from the project's Releases tab and copy it into the models directory:

   Image_Detector/models/custom_cnn_standalone.pt
   

---

Running the API Service

python app.py

The app will initialize and start listening on port 8000 by default.

Health Verification

Request:

curl http://localhost:8000/

Expected Response:

{
  "status": "healthy",
  "api_name": "Deep-Detect Image Classification Service",
  "model_architecture": "Custom CNN Standalone (PyTorch)",
  "device_running": "cpu",
  "endpoints": {
    "health_check": "/",
    "inference": "/predict"
  }
}

---

API Reference

Image Prediction Endpoint

• Endpoint: /predict
• Method: POST
• Payload format: multipart/form-data
• Field Name: file (must contain image data)

Success (200 OK)

{
  "prediction": "ai",
  "confidence": 94.85,
  "status": "success"
}

Bad Request (400)

{
  "detail": "Uploaded file must be a valid JPEG or PNG image."
}

---

Standalone Desktop Interface

If you want to perform predictions locally without running the web server daemon, you can run the Tkinter GUI tool:

python predict.py

This launches a graphical interface to browse, view, and inspect images.