Year: 2025
Type: conference-paper
Abstract: Deepfake technology, where artificial intelligence is used in order to generate highly realistic manipulated media, still is a severe threat to digital media authenticity. Detection of most deepfakes is really one of the most prominent computer vision tasks that we have at present. In this paper, we propose a specific deepfake detector. It is an attention and convolutional neural network (CNN). We employ an attention layer in our network to enable better salient detail representation along with trying to distinguish between forged and original images. Metrics including accuracy, precision, recall, and F1-score are measures used to determine the performance of the model after its training on a dataset of original and forged images. We additionally show the performance of the model by greatly improving the predictive outputs through face landmark detection and an image preprocessing pipeline.
Source: AIJR Proceedings
Institution: Vishwakarma Institute of Technology
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7127077361