Distill-then-prune: An Efficient Compression Framework for Real-time Stereo Matching Network on Edge Devices

TL;DR

This paper introduces a novel framework combining knowledge distillation and model pruning to create a lightweight, high-accuracy stereo matching network suitable for real-time edge device applications.

Contribution

It presents a new efficient compression strategy that effectively balances speed and accuracy for stereo matching networks, outperforming existing methods.

Findings

Achieved real-time performance with high accuracy on edge devices.

Demonstrated superior results on Sceneflow and KITTI benchmarks.

Validated the effectiveness of the distillation and pruning approach through extensive experiments.

Abstract

In recent years, numerous real-time stereo matching methods have been introduced, but they often lack accuracy. These methods attempt to improve accuracy by introducing new modules or integrating traditional methods. However, the improvements are only modest. In this paper, we propose a novel strategy by incorporating knowledge distillation and model pruning to overcome the inherent trade-off between speed and accuracy. As a result, we obtained a model that maintains real-time performance while delivering high accuracy on edge devices. Our proposed method involves three key steps. Firstly, we review state-of-the-art methods and design our lightweight model by removing redundant modules from those efficient models through a comparison of their contributions. Next, we leverage the efficient model as the teacher to distill knowledge into the lightweight model. Finally, we systematically prune the lightweight model to obtain the final model. Through extensive experiments conducted on two widely-used benchmarks, Sceneflow and KITTI, we perform ablation studies to analyze the effectiveness of each module and present our state-of-the-art results.