Improved ResNet-50 Model for AI Image Recognition Based on Multi-Scale Attention Mechanism
Chuyue Qi, Zonglin Yang, Yuxin Wen · 2024
Accurate recognition of AI-generated images is critical for information security and disinformation prevention in the media. Traditional image recognition models face challenges in recognizing images produced by advanced generation techniques, which may affect their accuracy and application scope. To address this issue, this paper proposes an improved ResNet-50 model, which significantly improves the ability to distinguish AI-generated images from real images by introducing a two-channel pooling layer and a multi-scale attention mechanism. The experimental results show that the model performs well in the image classification task with an accuracy of 99.13%, which is significantly higher than other comparative models. In addition, this study also conducted ablation experiments on the model to verify the respective contributions of the dual-channel pooling layer and the multi-scale attention mechanism to the performance improvement. These improvements make the model suitable for high-precision AI false image recognition tasks.