3rd Workshop on Maritime Computer Vision (MaCVi)

The workshop features five challenges designed to advance computer vision for autonomous boat navigation and underwater image restoration. Participants will work on a variety of tasks, such as object distance estimation, obstacle segmentation, and underwater image enhancement, with an emphasis on both accuracy and real-time operation on embedded hardware. The best three teams of each track will be featured as co-authors in a challenge summary workshop proceeding (included in IEEE Xplore) upon submitting a short technical report (max. half page) describing their method. We reserve the right to cancel a subchallenge if less than four separate teams participate. Note that only one model per team per track will count, and organizers may upload reference models that are not competing for prizes. Winners will be published in the WACV Workshop proceedings.

Quick links: Dataset downloads Submit Leaderboards Ask for help



Challenges Overview

Approximate Supervised Object Distance Estimation

Develop models for estimating object distances using monocular images captured from USVs - test
USV-based Obstacle Segmentation

Segment a maritime image from the viewpoint of a USV into three components (water, obstacle, sky)
USV-based Embedded Obstacle Segmentation

Segment a maritime image from the viewpoint of a USV in real-time on an embedded device
Prize: the top team stands to win $500 worth of Luxonis devices
USV-based Panoptic Segmentation

Parse the scene captured from the viewpoint of a USV into "stuff" classes and "thing" object instances
MarineVision Restoration Challenge

Restore degraded underwater images and detect aquatic animal species.