MPCNet: Improved MeshSegNet Based on Position Encoding and Channel Attention

In the process of orthodontic treatment, it is a very important step to accurately segment each tooth and jaw model with computer assistance. The use deep learning technology methods for segmentation can not only save lot manual interaction time cost but also improve treatment effect. 3D hot topic interest international related scholars, some end-to-end based on dental mesh scanning models have been emerging in recent years. Due limited variety existing models, they are well suited different scenarios, feature extraction capability effect these still need be improved. this paper, we propose novel method, MPCNet, which adds multi-scale density information input layer, uses position encoding channel attention mechanism MeshSegNet, graph-cut post-processing perform real scenes. effectiveness MPCNet evaluated scanned dataset, significantly outperforms current mainstream methods.