AMD: Adaptive Masked Distillation for Object Detection

Guang Yang, Yin Tang, Jun Li +2

As a general model compression paradigm, feature-based knowledge distillation allows the student model to learn expressive features from the teacher counterpart. In this paper, we…