Technical Report for ICRA 2025 GOOSE 2D Semantic Segmentation Challenge: Boosting Off-Road Segmentation via Photometric Distortion and Exponential Moving Average

TL;DR

This paper presents a high-capacity semantic segmentation approach for off-road environments, utilizing photometric distortion and EMA to improve generalization, achieving 88.8% mIoU in the GOOSE challenge.

Contribution

It demonstrates the effectiveness of combining established segmentation techniques with photometric augmentation and EMA in unstructured off-road scenes.

Findings

Achieved 88.8% mIoU on GOOSE validation set.

Photometric distortion improves robustness to lighting variations.

EMA of weights enhances model generalization.

Abstract

We report on the application of a high-capacity semantic segmentation pipeline to the GOOSE 2D Semantic Segmentation Challenge for unstructured off-road environments. Using a FlashInternImage-B backbone together with a UPerNet decoder, we adapt established techniques, rather than designing new ones, to the distinctive conditions of off-road scenes. Our training recipe couples strong photometric distortion augmentation (to emulate the wide lighting variations of outdoor terrain) with an Exponential Moving Average (EMA) of weights for better generalization. Using only the GOOSE training dataset, we achieve 88.8\% mIoU on the validation set.