You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

XAI-ForensiQ: Explainable AI Framework for Deepfake Detection

Model Architecture

Dual backbone: EfficientNet-B4 (17.5M) + ViT-B/16 (86M)
Total parameters: ~105M
Fusion: Linear(2560 → 512 → 1)
Best Val AUC: 0.9993 | Val Accuracy: 98.9%

Dataset

Celeb-DF v2 + FaceForensics++ C23 + 140k faces
Total: 170K+ images, balanced Real/Fake

XAI Components

Grad-CAM++ heatmaps (spatial localization)
SHAP analysis (channel-level attribution)
Gemini API forensic narrative generation

Setup

conda create -n deepfake python=3.10
conda activate deepfake
pip install -r requirements.txt

Repository Structure

models/ - trained model weights (.pth) scripts/ - training, evaluation, XAI scripts data/ - dataset CSVs (face images in dataset repo) outputs/ - sample GradCAM heatmaps + XAI reports

Face Images Dataset

Download from: https://huggingface.co/datasets/CrazyOp/XAI-ForensiQ-dataset

Usage

# Evaluate model
python scripts/evaluate.py

# Generate XAI report for an image
python scripts/gradcam_xai.py --image path/to/face.jpg

Citation

Coming soon — IEEE Access / Applied Sciences MDPI submission

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support