1st Workshop on Maritime Computer Vision (MaCVi)
An integral part of the workshop are the four challenges UAV Object Detection & Tracking and USV Obstacle Segmentation & Detection. These challenges aim to
foster the development of computer vision algorithms in SaR missions and autonomous boat navigation, respectively. The UAV Object Detection & Tracking challenge is mainly based on
an updated version of SeaDronesSee whereas the USV Obstacle Segmentation & Detection is based on the MODS benchmark.
The focus of all challenges will be on accuracy (algorithm performance as measured by challenge-specific evaluation metrics),
but require participants to submit the processing speed of their algorithms in frames per second, and specification of the hardware used.
Top performing teams of each track (Binary OD v2 counts as a separate track) will be featured as co-authors in a challenge summary workshop proceeding (included in IEEE Xplore) upon submitting a short technical report (half page) describing their method. We reserve the right to cancel a subchallenge if less than four separate teams (no collusion allowed) participate. Note that only one model per team per track counts even if multiple models of the same team are within the top 3, i.e. the worse models will be ignored in the leaderboard (but do not hurt). You may participate and win in multiple tracks still. Also, note the short deadline for the technical report.
Note that we (as organizers) may upload models for the challenges BUT we do not compete for a winning position (i.e. our models do not count on the leaderboard and merely serve as references). Thus, if your method is worse (in any metric) than one of the organizer's, you are still encouraged to submit your method as you might win.
Lastly, Sentient Vision Systems are happy to sponsor the UAV-based Object Detection v2 challenge track. Each of the three winning teams will receive a GPU card selected and supplied by Sentient (details soon).
Top performing teams of each track (Binary OD v2 counts as a separate track) will be featured as co-authors in a challenge summary workshop proceeding (included in IEEE Xplore) upon submitting a short technical report (half page) describing their method. We reserve the right to cancel a subchallenge if less than four separate teams (no collusion allowed) participate. Note that only one model per team per track counts even if multiple models of the same team are within the top 3, i.e. the worse models will be ignored in the leaderboard (but do not hurt). You may participate and win in multiple tracks still. Also, note the short deadline for the technical report.
Note that we (as organizers) may upload models for the challenges BUT we do not compete for a winning position (i.e. our models do not count on the leaderboard and merely serve as references). Thus, if your method is worse (in any metric) than one of the organizer's, you are still encouraged to submit your method as you might win.
Lastly, Sentient Vision Systems are happy to sponsor the UAV-based Object Detection v2 challenge track. Each of the three winning teams will receive a GPU card selected and supplied by Sentient (details soon).
Challenges Overview
 |
UAV Object Detection v2 Detect humans, boats and others in a search and rescue scenario from the viewpoint of a UAV |
|---|---|
 |
UAV Multi-Object Tracking Track detected objects in a search and rescue scenario from the viewpoint of a UAV |
 |
USV Obstacle Segmentation Segment a maritime image from the viewpoint of a USV into three components (water, obstacle, sky) |
 |
USV Obstacle Detection Detect dynamic obstacles like ships, swimmers, boats and buoys on a maritime image from the viewpoint of a USV |
