VisionShield is an ongoing research-driven project exploring the use of deep learning and computer vision techniques to detect AI-generated (DeepFake) media. The system integrates convolutional neural networks (CNNs) with OpenCV-based preprocessing to identify subtle facial artifacts and inconsistencies that distinguish synthetic faces from authentic ones.
- Develop a hybrid Neural Network–OpenCV framework capable of detecting DeepFake manipulations in real time.
- Utilize benchmark datasets such as FaceForensics++ and DeepFake Detection Challenge (DFDC) for model training and evaluation.
- Apply Grad-CAM visualization to improve explainability by highlighting the facial regions influencing model predictions.
- Implement a Streamlit-based interface for real-time inference and result interpretation.
This project aims to contribute to the growing field of AI for digital media forensics, emphasizing model interpretability, adversarial robustness, and responsible use of generative AI technologies.