First Place Solution to the ECCV 2024 BRAVO Challenge: Evaluating Robustness of Vision Foundation Models for Semantic Segmentation

Tommie Kerssies, Daan de Geus, Gijs Dubbelman
9/25/2024
cs.CVcs.AIcs.LG

Official code: tue-mps/benchmark-vfm-ss (38 stars, MIT, last commit 2025-07-18).

Abstract

In this report, we present the first place solution to the ECCV 2024 BRAVO Challenge, where a model is trained on Cityscapes and its robustness is evaluated on several out-of-distribution datasets. Our solution leverages the powerful representations learned by vision foundation models, by attaching a simple segmentation decoder to DINOv2 and fine-tuning the entire model. This approach outperforms more complex existing approaches, and achieves first place in the challenge. Our code is publicly available at https://github.com/tue-mps/benchmark-vfm-ss.

View on arXivView PDF

Code Implementations(1)

tue-mps/benchmark-vfm-ssOfficial100%
MIT
384Python, Jupyter NotebookApr 9, 20241 years agoMIT
benchmarkfoundation-modelsemantic-segmentationvfmvision-foundation-model+1 more

Similar Papers

How to Benchmark Vision Foundation Models for Semantic Segmentation?

84% match1 repo382024

The BRAVO Semantic Segmentation Challenge Results in UNCV2024

84% match1 repo212024

The Robust Semantic Segmentation UNCV2023 Challenge Results

82% match2023

A Novel Benchmark for Few-Shot Semantic Segmentation in the Era of Foundation Models

81% match2 repos112024

Technical Report for CVPR 2024 WeatherProof Dataset Challenge: Semantic Segmentation on Paired Real Data

81% match2024

Revealing the Semantic Selection Gap in DINOv3 through Training-Free Few-Shot Segmentation

80% match1 repo42026

Cite this paper

@article{kerssies2024first,
  title  = {First Place Solution to the ECCV 2024 BRAVO Challenge: Evaluating Robustness of Vision Foundation Models for Semantic Segmentation},
  author = {Tommie Kerssies and Daan de Geus and Gijs Dubbelman},
  year   = {2024},
  eprint = {2409.17208},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2409.17208v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2409.17208.svg)](https://codeofpaper.com/paper/2409.17208)

Discussion