Knowledge Distillation (KD) is a well-known training paradigm in deep neural networks where knowledge acquired by a large teacher model is transferred to a small student. KD has proven to be an effective technique to significantly improve the student's performance for various tasks including object detection. As such, KD techniques mostly rely on guidance at the intermediate feature level, which is typically implemented by minimizing an L_p-norm distance between teacher and student activations during training. In this paper, we propose a replacement for the pixel-wise independent L_p-norm based on structural similarity (SSIM). By taking into account additional contrast and structural cues, more knowledge within intermediate feature maps can be preserved. Extensive experiments on MSCOCO demonstrate the effectiveness of our method across different training schemes and architectures.
Our method adds only little computational overhead, is straightforward to implement and at the same time it significantly outperforms the standard L_p-norms. Moreover, more complex state-of-the-art KD methods using attention-based sampling mechanisms are outperformed, including a +3.5 AP gain using a Faster R-CNN R-50 compared to a vanilla model.