🧠 MesoNet Trained Model (Deepfake Detection)

This repository contains the trained weights of a MesoNet-based deepfake detection model, saved in .h5 format. The model was trained on the UADFV_dataset containing both real and fake face videos.

---

📁 Files in this Repository

File	Description
mesonet_model.h5	Trained MesoNet model in Keras HDF5 format

---

🧠 Model Overview: MesoNet

• Lightweight CNN for detecting facial forgery in images
• Designed to detect subtle inconsistencies at the mesoscopic level
• Works on 256×256 facial frames
• Outputs a probability score for classification: Real or Fake

Based on the paper: MesoNet: a Compact Facial Video Forgery Detection Network

---

📊 Dataset Used

• Dataset Name: UADFV_dataset
• Source: Kaggle – UADFV_dataset
• Includes real and deepfake videos
• Frames extracted and resized to 256×256 for training

---

🚀 How to Use

1. Clone this repo or download the .h5 file directly:

git clone https://github.com/muffin-123/mesonet-trained-model.git
cd mesonet-trained-model