LiteSeg: A Novel Lightweight ConvNet for Semantic Segmentation

TL;DR

LiteSeg is a new lightweight convolutional neural network designed for semantic segmentation, combining depthwise separable convolutions, residual connections, and an enhanced ASPP module to achieve high accuracy and speed.

Contribution

The paper introduces LiteSeg, a novel lightweight architecture that improves efficiency and accuracy in semantic segmentation using innovative modules and multiple backbone networks.

Findings

Achieves 67.81% mIoU on Cityscapes with MobileNetV2 backbone.

Operates at 161 frames per second at 640x360 resolution.

Demonstrates effective trade-offs between accuracy and computational cost.

Abstract

Semantic image segmentation plays a pivotal role in many vision applications including autonomous driving and medical image analysis. Most of the former approaches move towards enhancing the performance in terms of accuracy with a little awareness of computational efficiency. In this paper, we introduce LiteSeg, a lightweight architecture for semantic image segmentation. In this work, we explore a new deeper version of Atrous Spatial Pyramid Pooling module (ASPP) and apply short and long residual connections, and depthwise separable convolution, resulting in a faster and efficient model. LiteSeg architecture is introduced and tested with multiple backbone networks as Darknet19, MobileNet, and ShuffleNet to provide multiple trade-offs between accuracy and computational cost. The proposed model LiteSeg, with MobileNetV2 as a backbone network, achieves an accuracy of 67.81% mean intersection over union at 161 frames per second with $640 \times 360$ resolution on the Cityscapes dataset.