Reparameterizable Dual-Resolution Network for Real-time Semantic Segmentation

TL;DR

This paper introduces RDRNet, a dual-resolution network with reparameterization techniques that improve real-time semantic segmentation accuracy and speed, suitable for applications like autonomous driving.

Contribution

The study proposes a novel reparameterizable dual-resolution architecture and a pyramid pooling module that enhance segmentation performance without increasing inference time.

Findings

Outperforms state-of-the-art models on Cityscapes, CamVid, and Pascal VOC 2012 datasets.

Achieves a better balance of accuracy and inference speed.

Demonstrates the effectiveness of reparameterization in real-time segmentation.

Abstract

Semantic segmentation plays a key role in applications such as autonomous driving and medical image. Although existing real-time semantic segmentation models achieve a commendable balance between accuracy and speed, their multi-path blocks still affect overall speed. To address this issue, this study proposes a Reparameterizable Dual-Resolution Network (RDRNet) dedicated to real-time semantic segmentation. Specifically, RDRNet employs a two-branch architecture, utilizing multi-path blocks during training and reparameterizing them into single-path blocks during inference, thereby enhancing both accuracy and inference speed simultaneously. Furthermore, we propose the Reparameterizable Pyramid Pooling Module (RPPM) to enhance the feature representation of the pyramid pooling module without increasing its inference time. Experimental results on the Cityscapes, CamVid, and Pascal VOC 2012 datasets demonstrate that RDRNet outperforms existing state-of-the-art models in terms of both performance and speed. The code is available at https://github.com/gyyang23/RDRNet.