Smooth and Stepwise Self-Distillation for Object Detection

TL;DR

This paper introduces Smooth and Stepwise Self-Distillation (SSSD), a novel method that improves object detection by using a smoother Jensen-Shannon based loss and adaptive distillation coefficients, outperforming prior methods.

Contribution

The paper proposes SSSD, a new self-distillation framework for object detection that employs a smoother Jensen-Shannon loss and adaptive coefficients, enhancing robustness and accuracy.

Findings

SSSD achieves higher average precision than baselines and state-of-the-art detectors.

SSSD is robust across various coefficient settings and backbone architectures.

Stepwise distillation significantly improves detection performance.

Abstract

Distilling the structured information captured in feature maps has contributed to improved results for object detection tasks, but requires careful selection of baseline architectures and substantial pre-training. Self-distillation addresses these limitations and has recently achieved state-of-the-art performance for object detection despite making several simplifying architectural assumptions. Building on this work, we propose Smooth and Stepwise Self-Distillation (SSSD) for object detection. Our SSSD architecture forms an implicit teacher from object labels and a feature pyramid network backbone to distill label-annotated feature maps using Jensen-Shannon distance, which is smoother than distillation losses used in prior work. We additionally add a distillation coefficient that is adaptively configured based on the learning rate. We extensively benchmark SSSD against a baseline and two state-of-the-art object detector architectures on the COCO dataset by varying the coefficients and backbone and detector networks. We demonstrate that SSSD achieves higher average precision in most experimental settings, is robust to a wide range of coefficients, and benefits from our stepwise distillation procedure.