SRC USAGE INSTRUCTIONS

DOWNLOAD BOTH MODELS FIRST BEFORE RUNNING CODE, THEN PROCEED WITH THE NEXT STEPS Download link: Google Drive

1. Open terminal (Switch to Git Bash) and check the current directory

• If in /src, no action needed
• If in /project, run cd src

2. Run Commands

FOR FIRST TIME SETUP, RUN STEP 2.1

• 2.1. python -m venv venv (Create virtual environment - creates a venv folder in src)
• 2.2. source venv/Scripts/activate (Activate virtual environment - .venv will appear in the prompt)
• 2.3. pip install -r requirements.txt (Install required packages - if an error occurs, check the path to requirements.txt)
• 2.4. python app.py (Run this inside the terminal with virtual environment activated)

If started successfully, the Flask log will appear. Open your browser at: http://127.0.0.1:5000/

========

Key Features

Image Upload

Select an image from your device → the system will:

Run YOLO for object detection.

Run custom EfficientNet-B4 for REAL/FAKE classification.

Display original image, image with bounding boxes, YOLO results table, and the REAL/FAKE verdict.

Realtime Webcam

Open webcam → the system will:

Detect only the person class.

Overlay REAL/FAKE result + person label directly onto the video feed.

Display 2 lines of information below the video:

REAL/FAKE verdict

Detected entity (person / no person detected)

========

PROJECT STRUCTURE

src/ │ ├── app.py # Main Flask server ├── templates/ │ ├── index.html # Image upload interface │ └── webcam.html # Realtime webcam interface ├── static/ │ └── uploads/ # Uploaded images and YOLO results ├── requirements.txt # Required Python packages ├── best_face_fake_detector14b211.pth # Custom EfficientNet-B4 checkpoint └── yolo11n.pt # YOLO model