Image Tampering Recognition Algorithm Based on Improved YOLOv5s

In order to improve the extraction effect of features tampered region, model usually relies on a specific feature defined by hand optimize model. However, is not ideal when mode unknown. this paper, Neck layer YOLOv5s reasonably embedded into CBAM attention module, so that can more accurately capture region. addition, boundary frame loss optimized using EIOU function, which separately dissolves impact factors aspect ratio calculate The experimental results show image tampering recognition algorithm proposed in paper based improved only effectively identify variety modes, but also increase average accuracy 1.57% compared with benchmark algorithm. speed higher than LSTM and U-Net algorithms, reaches 13.89 images per second, maintaining high level.