Lightweight Alpha Matting Network Using Distillation-Based Channel Pruning

Recently, alpha matting has received a lot of attention because its usefulness in mobile applications such as selfies. Therefore, there been demand for lightweight model due to the limited computational resources commercial portable devices. To this end, we suggest distillation-based channel pruning method networks. In step, remove channels student network having fewer impacts on mimicking knowledge teacher network. Then, pruned is trained by same distillation loss. A from proposed outperforms existing methods. show superiority our algorithm, provide various quantitative and qualitative experiments with in-depth analyses. Furthermore, demonstrate versatility applying it semantic segmentation.