An Image Forgery Detection Technology Based on Deep Learning
Yuqiang Zou, Junzhu Liu, Yicong Li, Yuxuan Xie, Bocheng Liu, Tingfeng Yi, Xuan Li · 2024
As artificial intelligence advances, deep learning-based facial forgery techniques have become increasingly sophisticated, making it challenging for the human eye to distinguish these fake facial images. To prevent malicious exploitation and potential societal panic, there is a crucial need for the development of efficient and reliable facial forgery detection technology. Current deep learning-based detection models suffer from poor generality and inadequate facial feature extraction, resulting in low accuracy. This study enhances an Xception-based model through transfer learning, improving training speed without compromising accuracy. Comparative analysis shows that the designed F-Xception model significantly enhances robustness while maintaining a rapid convergence speed compared to Xception and MesoNet networks on various datasets.