Saved in:
Bibliographic Details
Main Authors: Toledo, Rafael S., Oliveira, Cristiano S., Oliveira, Vitor H. T., Antonelo, Eric A., von Wangenheim, Aldo
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.16295
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909403843657728
author Toledo, Rafael S.
Oliveira, Cristiano S.
Oliveira, Vitor H. T.
Antonelo, Eric A.
von Wangenheim, Aldo
author_facet Toledo, Rafael S.
Oliveira, Cristiano S.
Oliveira, Vitor H. T.
Antonelo, Eric A.
von Wangenheim, Aldo
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2411_16295
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle A Performance Increment Strategy for Semantic Segmentation of Low-Resolution Images from Damaged Roads
Toledo, Rafael S.
Oliveira, Cristiano S.
Oliveira, Vitor H. T.
Antonelo, Eric A.
von Wangenheim, Aldo
Computer Vision and Pattern Recognition
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.
title A Performance Increment Strategy for Semantic Segmentation of Low-Resolution Images from Damaged Roads
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.16295