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

TL;DR

This work proposes the Performance Increment Strategy for Semantic Segmentation (PISSS) as a methodology of 14 training experiments to boost performance and offers an analysis of DeepLabV3+ pitfalls for small object segmentation.

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.

Figures (8)

• Figure 1: Examples of emerging countries' roads in the RTK dataset.
• Figure 2: PISSS Diagram.
• Figure 3: mIoU oscillation after loss convergence.
• Figure 4: Low and high-level features connections in DeepLabV3+.
• Figure 5: Comparing results of distinct OS w/ or wo/ MP layer. The subtitles have the OS and MP states and the prediction IoU result. Removing the MP layer avoids early spatial information loss for extracting small object features.