Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

3rd Workshop on Maritime Computer Vision (MaCVi) 2025: Challenge Results

Benjamin Kiefer, Lojze Žust, Jon Muhovič, Matej Kristan, Janez Perš, Matija Teršek, Uma Mudenagudi Chaitra Desai, Arnold Wiliem, Marten Kreis, Nikhil Akalwadi, Yitong Quan, Zhiqiang Zhong, Zhe Zhang, Sujie Liu, Xuran Chen, Yang Yang, Matej Fabijanić, Fausto Ferreira, Seongju Lee, Junseok Lee, Kyoobin Lee, Shanliang Yao, Runwei Guan, Xiaoyu Huang, Yi Ni, Himanshu Kumar, Yuan Feng, Yi-Ching Cheng, Tzu-Yu Lin, Chia-Ming Lee, Chih-Chung Hsu, Jannik Sheikh, Andreas Michel, Wolfgang Gross, Martin Weinmann, Josip Šarić, Yipeng Lin, Xiang Yang, Nan Jiang, Yutang Lu, Fei Feng, Ali Awad, Evan Lucas, Ashraf Saleem, Ching-Heng Cheng, Yu-Fan Lin, Tzu-Yu Lin, Chih-Chung Hsu

TL;DR

MaCVi 2025 presents a comprehensive suite of maritime vision challenges for USVs and UUVs, covering distance estimation, semantic and panoptic segmentation (including embedded deployment), and underwater image restoration. The methods span detector-centric augmentations (BiFormer, SimAM), lightweight embedded architectures, and open-vocabulary approaches, with LaRS-based benchmarks driving evaluation for real-world navigation, obstacle avoidance, and robustness to lighting and reflections. Key contributions include competitive distance estimation strategies with metric reasoning, state-of-the-art segmentation performance on LaRS across regular and embedded settings, and a two-stage underwater restoration-plus-detection pipeline that surpasses baselines in marine species detection. Collectively, the results demonstrate meaningful gains in real-time maritime perception, industrial-grade benchmarking, and open research avenues for multi-task, open-world, and hardware-constrained scenarios with direct implications for autonomous surface and underwater vehicles.

Abstract

The 3rd Workshop on Maritime Computer Vision (MaCVi) 2025 addresses maritime computer vision for Unmanned Surface Vehicles (USV) and underwater. This report offers a comprehensive overview of the findings from the challenges. We provide both statistical and qualitative analyses, evaluating trends from over 700 submissions. All datasets, evaluation code, and the leaderboard are available to the public at https://macvi.org/workshop/macvi25.

3rd Workshop on Maritime Computer Vision (MaCVi) 2025: Challenge Results

TL;DR

MaCVi 2025 presents a comprehensive suite of maritime vision challenges for USVs and UUVs, covering distance estimation, semantic and panoptic segmentation (including embedded deployment), and underwater image restoration. The methods span detector-centric augmentations (BiFormer, SimAM), lightweight embedded architectures, and open-vocabulary approaches, with LaRS-based benchmarks driving evaluation for real-world navigation, obstacle avoidance, and robustness to lighting and reflections. Key contributions include competitive distance estimation strategies with metric reasoning, state-of-the-art segmentation performance on LaRS across regular and embedded settings, and a two-stage underwater restoration-plus-detection pipeline that surpasses baselines in marine species detection. Collectively, the results demonstrate meaningful gains in real-time maritime perception, industrial-grade benchmarking, and open research avenues for multi-task, open-world, and hardware-constrained scenarios with direct implications for autonomous surface and underwater vehicles.

Abstract

The 3rd Workshop on Maritime Computer Vision (MaCVi) 2025 addresses maritime computer vision for Unmanned Surface Vehicles (USV) and underwater. This report offers a comprehensive overview of the findings from the challenges. We provide both statistical and qualitative analyses, evaluating trends from over 700 submissions. All datasets, evaluation code, and the leaderboard are available to the public at https://macvi.org/workshop/macvi25.
Paper Structure (60 sections, 4 equations, 17 figures, 6 tables)

This paper contains 60 sections, 4 equations, 17 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Challenge Participation Protocol
  3. Submission Process
  4. Evaluation Server and Leaderboard
  5. Timeline
  6. Recognition and Rewards
  7. USV-based Perception Challenges
  8. Approximate Supervised Distance Estimation
  9. Evaluation Protocol
  10. Submissions, Analysis and Trends
  11. Discussion and Challenge Winners
  12. USV-based Obstacle Segmentation Challenge
  13. Evaluation Protocol
  14. Submissions, Analysis and Trends
  15. Discussion and Challenge Winners
  16. ...and 45 more sections

Figures (17)

  • Figure 1: Overview of MaCVi2025 challenges. Challenges include USV-based (a) Distance Estimation, (b) Regular and Embedded Obstacle Segmentation, (c) Panoptic Segmentation, and UUV-based (d) Image Restoration.
  • Figure 2: Distribution of samples in the training () and test () splits with respect to the ground truth distance
  • Figure 3: Common buoy types in the supervised distance estimation dataset
  • Figure 4: Mean IoU over the distance to the objects with a step size of 100 meters for the top 2 submissions compared with the baseline model
  • Figure 5: Distance Estimation Results comparing the top two submissions, Data Enhance and Yolov7 Depth Widened, against the provided baseline model with the absolute error metric ($\epsilon_{abs}$).
  • ...and 12 more figures