Stack of Heatmap Hourglass Network for Face Alignment

This paper presents a new approach for facial landmark localization using a stack of hourglass models and an attention mechanism. The proposed method utilizes heatmap regression and a multi-stage architecture composed of hourglass networks to locate key points on a face, such as the eyes, eyebrows, nose, and mouth. The attention model focuses the network’s attention on specific features, enabling more accurate localization of facial landmarks. The model is trained on a largescale dataset using the Mean Squared Error loss function and evaluated on benchmark datasets. The results demonstrate that the proposed method achieves state-of-the-art performance on three popular benchmark datasets, significantly improving the Normalized Mean Error metric (NME). The highest NME was 2.81 on the 300W data set. Fig. 1 shows some of our results. Furthermore, the method shows robustness to pose, illumination, and occlusion variations, making it a promising approach for real-world applications in face recognition, Face Age estimation, and facial expression analysis. The proposed method offers a promising solution for accurate and robust facial landmark localization in computer vision.