(ii) We propose a Fast Pose Distillation (FPD) model training method enabling to more effectively train ex- tremely small human pose CNN networks. This is based ...
確定! 回上一頁