Multi-Scale Aligned Distillation for Low-Resolution Detection

TL;DR

This paper introduces a novel multi-scale aligned distillation method that enhances low-resolution object detection models by effectively transferring knowledge from multi-resolution teachers, significantly improving their accuracy.

Contribution

It proposes a spatially aligned multi-scale distillation framework with feature fusion, enabling low-resolution models to achieve high performance comparable to high-resolution models.

Findings

Low-resolution models trained with our method outperform traditional low-res models by 2.1% to 3.6% mAP.

Our approach achieves competitive results with high-resolution models using low-resolution inputs.

The method effectively aligns features across different resolutions for improved knowledge transfer.

Abstract

In instance-level detection tasks (e.g., object detection), reducing input resolution is an easy option to improve runtime efficiency. However, this option traditionally hurts the detection performance much. This paper focuses on boosting the performance of low-resolution models by distilling knowledge from a high- or multi-resolution model. We first identify the challenge of applying knowledge distillation (KD) to teacher and student networks that act on different input resolutions. To tackle it, we explore the idea of spatially aligning feature maps between models of varying input resolutions by shifting feature pyramid positions and introduce aligned multi-scale training to train a multi-scale teacher that can distill its knowledge to a low-resolution student. Further, we propose crossing feature-level fusion to dynamically fuse teacher's multi-resolution features to guide the student better. On several instance-level detection tasks and datasets, the low-resolution models trained via our approach perform competitively with high-resolution models trained via conventional multi-scale training, while outperforming the latter's low-resolution models by 2.1% to 3.6% in terms of mAP. Our code is made publicly available at https://github.com/dvlab-research/MSAD.