3rd Workshop on Maritime Computer Vision (MaCVi) 2025: Challenge Results
🤖 AI Summary
Computer vision models for unmanned surface vessels (USVs) and underwater scenarios suffer from poor robustness, compounded by the absence of standardized evaluation protocols. Method: The MaCVi 2025 workshop organized a global competition to systematically benchmark over 700 technical submissions, establishing the first unified maritime computer vision evaluation framework. This framework integrates statistical analysis with qualitative assessment across core tasks—including object detection, image segmentation, and multimodal perception—and leverages publicly available datasets, standardized evaluation code, and a real-time leaderboard to ensure reproducibility. Contribution/Results: The initiative delivers the first open-source, authoritative maritime vision benchmark—comprising all data, reference implementations, and a live leaderboard—significantly enhancing model generalization and robustness in challenging aquatic environments (e.g., turbid water, dynamic lighting, occlusion). It lays the foundational infrastructure for standardization and advancement of maritime AI research.
Application Category
📝 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.
Problem
Research questions and friction points this paper is trying to address.
Maritime Computer Vision
Unmanned Vessels
Underwater Technology
Innovation
Methods, ideas, or system contributions that make the work stand out.
Marine Computer Vision
Unmanned Marine Vessels
Underwater Environments
🔎 Similar Papers
No similar papers found.
