Face Swap Detection using Deep Learning Techniques

Dr.M.V.Krishna Mohan,Associate Professor
Ragolu Vineetha, Vindula Devi Priyanka, Pappu Siva Sai Prasad, Padala sai venkata Prasad
Department of Computer Science and Engineering
Visakha Institute of Engineering and Technology, Visakhapatnam, Andhra Pradesh, India

Abstract:Face swap and deepfake manipulations have become a serious threat to the authenticity of digital images and videos.Such manipulations can be misused for spreading misinformation, identity fraud, and damaging public trust. Manual checking is slow and not suitable for large-scale use. This paper presents a deep learning-based system for automatic face swap detection. The system uses OpenCV for face detection and alignment combined with a fine-tuned ResNet50 model for classification. It distinguishes real faces from swapped faces and achieves a validation accuracy of 98.75%. Gradient-weighted Class Activation Mapping (Grad-CAM) is integrated to highlight manipulated regions such as blending artifacts and texture inconsistencies, making the results more understandable. The model is deployed as a simple web application that provides real-time predictions, confidence scores, visual explanations, and useful recommendations. The system also supports multiple languages to improve accessibility. Experimental results show that the proposed approach is accurate, interpretable, and practical for real-world digital forensics and content moderation applications.Keywords:Face Swap Detection, Deepfake Detection, Deep Learning, ResNet50, OpenCV, Grad-CAM, Digital Forensics