Distilling Object Detectors with Feature Richness

TL;DR

This paper introduces a novel feature-richness score method for distilling object detectors, effectively selecting important features outside bounding boxes and removing detrimental ones, leading to improved small model performance.

Contribution

The paper proposes the Feature-Richness Score (FRS) method for better feature selection during distillation, enhancing small detector accuracy beyond existing methods.

Findings

Achieves 39.7% mAP on COCO2017 with RetinaNet and ResNet-50.

Surpasses the teacher detector performance with ResNet-101.

Effective for both anchor-based and anchor-free detectors.

Abstract

In recent years, large-scale deep models have achieved great success, but the huge computational complexity and massive storage requirements make it a great challenge to deploy them in resource-limited devices. As a model compression and acceleration method, knowledge distillation effectively improves the performance of small models by transferring the dark knowledge from the teacher detector. However, most of the existing distillation-based detection methods mainly imitating features near bounding boxes, which suffer from two limitations. First, they ignore the beneficial features outside the bounding boxes. Second, these methods imitate some features which are mistakenly regarded as the background by the teacher detector. To address the above issues, we propose a novel Feature-Richness Score (FRS) method to choose important features that improve generalized detectability during distilling. The proposed method effectively retrieves the important features outside the bounding boxes and removes the detrimental features within the bounding boxes. Extensive experiments show that our methods achieve excellent performance on both anchor-based and anchor-free detectors. For example, RetinaNet with ResNet-50 achieves 39.7% in mAP on the COCO2017 dataset, which even surpasses the ResNet-101 based teacher detector 38.9% by 0.8%. Our implementation is available at https://github.com/duzhixing/FRS.