Physiological Curves Extraction of Human Ear Based on Improved YOLACT

In the field of ear shape clustering, 3D ear modeling and ear related personal product implementation, it is of great significance to retrieve the accurate position of some key physiological curves and key points. Traditional edge extraction method is very sensitive to lighting and pose variations. A method for extracting the key physiological curves of the human ear based on an improved instance segmentation network YOLACT is proposed in this paper. By labeling the human ear dataset, the YOLACT model is trained. By improving the network prediction process, it can accurately segment different regions of the human ear and extract key physiological curves. Compared with the original YOLACT and other segmentation methods, it shows higher segmentation accuracy on the test dataset and can extract more accurate human ear physiological curves. The proposed method also shows more robustness under lighting variation, pose variation and slight occlusion.