Distilling Object Detectors with Task Adaptive Regularization

TL;DR

This paper introduces a task adaptive knowledge distillation framework for object detectors, selectively transferring knowledge to improve efficiency and accuracy, especially on low-end devices.

Contribution

It proposes a novel adaptive distillation method that transfers knowledge based on task-specific priors and introduces a distillation decay strategy to enhance model generalization.

Findings

Achieves 39.0% mAP on COCO with ResNet-50, surpassing the baseline and even the teacher model.

Effectively transfers knowledge at multiple levels of the detector architecture.

Demonstrates improved detection performance with reduced computational costs.

Abstract

Current state-of-the-art object detectors are at the expense of high computational costs and are hard to deploy to low-end devices. Knowledge distillation, which aims at training a smaller student network by transferring knowledge from a larger teacher model, is one of the promising solutions for model miniaturization. In this paper, we investigate each module of a typical detector in depth, and propose a general distillation framework that adaptively transfers knowledge from teacher to student according to the task specific priors. The intuition is that simply distilling all information from teacher to student is not advisable, instead we should only borrow priors from the teacher model where the student cannot perform well. Towards this goal, we propose a region proposal sharing mechanism to interflow region responses between the teacher and student models. Based on this, we adaptively transfer knowledge at three levels, \emph{i.e.}, feature backbone, classification head, and bounding box regression head, according to which model performs more reasonably. Furthermore, considering that it would introduce optimization dilemma when minimizing distillation loss and detection loss simultaneously, we propose a distillation decay strategy to help improve model generalization via gradually reducing the distillation penalty. Experiments on widely used detection benchmarks demonstrate the effectiveness of our method. In particular, using Faster R-CNN with FPN as an instantiation, we achieve an accuracy of $39.0\%$ with Resnet-50 on COCO dataset, which surpasses the baseline $36.3\%$ by $2.7\%$ points, and even better than the teacher model with $38.5\%$ mAP.