A Performance Increment Strategy for Semantic Segmentation of Low-Resolution Images from Damaged Roads

TL;DR

This paper introduces PISSS, a strategy of 14 training experiments that significantly improves semantic segmentation performance on low-resolution, damaged road images, addressing unique challenges in emerging countries.

Contribution

The paper proposes PISSS, a novel methodology with 14 experiments, to enhance semantic segmentation of low-res damaged roads, achieving state-of-the-art results.

Findings

Achieved 79.8 mIoU on RTK dataset.

Achieved 68.8 mIoU on TAS500 dataset.

Analyzed DeepLabV3+ limitations for small objects.

Abstract

Autonomous driving needs good roads, but 85% of Brazilian roads have damages that deep learning models may not regard as most semantic segmentation datasets for autonomous driving are high-resolution images of well-maintained urban roads. A representative dataset for emerging countries consists of low-resolution images of poorly maintained roads and includes labels of damage classes; in this scenario, three challenges arise: objects with few pixels, objects with undefined shapes, and highly underrepresented classes. To tackle these challenges, this work proposes the Performance Increment Strategy for Semantic Segmentation (PISSS) as a methodology of 14 training experiments to boost performance. With PISSS, we reached state-of-the-art results of 79.8 and 68.8 mIoU on the Road Traversing Knowledge (RTK) and Technik Autonomer Systeme 500 (TAS500) test sets, respectively. Furthermore, we also offer an analysis of DeepLabV3+ pitfalls for small object segmentation.