Facial Landmark Detection Algorithm Based on an Improved Residual Self-Attention Mechanism

In recent years, facial landmark detection based on deep learning has made significant progress. However, it remains a challenging task, involving issues such as increased computational overhead and the risk of overfitting due to larger model parameters—especially when facial images suffer from occlusion or exhibit significant expression variations. To address these challenges, this paper proposes a facial landmark detection algorithm based on an improved residual self-attention mechanism network. Experimental results show that the proposed method is better than other algorithms on the public 300W dataset.