Image Tampering Recognition Algorithm Based on Improved YOLOv5s

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