Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{201467,
        author = {Manepalli Santhosh Babu and Dr. P. Sumalatha},
        title = {A Deep Learning-Based Approach for Deepfake Image and Video Detection Using ResNet50},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {3958-3965},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=201467},
        abstract = {The proliferation of synthetic media generated through Generative Adversarial Networks (GANs) has created urgent challenges for digital forensics, cybersecurity, and information integrity. Deepfakes—photo-realistic manipulated images and videos—have advanced to a degree where human perception alone is insufficient for reliable detection. This paper presents a deep learning-based deepfake detection system employing a fine-tuned ResNet50 Convolutional Neural Network (CNN) for binary classification of media as real or fake. The proposed system handles both static images and video sequences through a frame-sampling pipeline followed by temporal aggregation. A strict decision threshold of 55% is applied: media is classified as fake if the model's confidence score meets or exceeds 55%, and as real otherwise. To ensure transparency and interpretability, Gradient-weighted Class Activation Mapping (Grad-CAM) is integrated to generate spatial heatmaps that visually explain which regions of an image drive the model's decision. The system is implemented using Python, PyTorch, and OpenCV, and deployed via a Gradio-based web interface. Experimental evaluation on the FaceForensics++ and Celeb-DF datasets demonstrates an overall accuracy of 92.5%, precision of 91.2%, recall of 90.8%, and an F1-score of 91.0%. These results establish the viability of combining transfer learning with explainable AI for robust, interpretable deepfake detection.},
        keywords = {Deepfake Detection, Convolutional Neural Networks, ResNet50, Transfer Learning, Grad-CAM, Explainable AI, Video Forensics, Binary Classification.},
        month = {May},
        }