xView3: Prize Challenge to Detect Illegal Fishing on Synthetic Aperture Radar Imagery

Hello fast.ai community!

Our team at the Defense Innovation Unit, Global Fishing Watch, US Coast Guard, and NOAA are proud to announce the xView3 computer vision prize challenge to detect acts of illegal, unreported, and unregulated (IUU) fishing on synthetic aperture radar (SAR) imagery. With a prize pool of $150,000, xView3 will be the largest dark vessel dataset in the world.

The challenge press release is here: https://www.diu.mil/latest/us-government-and-nonprofit-organization-host-prize-competition-xview3
You can pre-register for the challenge here: https://www.diu.mil/ai-xview-challenge#xview3

Please do let me know if you have any questions!


Prizes only for US residents

Hi! I was just about to open a post about this. On that note, to you or anyone who may know: any idea what the SotA benchmarks on SAR detection/segmentation are? I’ll update if I find them, but it isn’t as straight forward as imagenet or coco.

(2020) Introduced via SpaceNet 6: MSAW dataset (0.12-0.14 F1 baseline)

(2020) Deep Learning Meets SAR – gives an overview of datasets

(2017) Kaggle Statoil/C-CORE Iceberg Classifier Challenge – leaderboard could serve as a proof of concept benchmark.

Yes and No :slight_smile:

One part of awards is really for US residents only, however other one:

“Global Fishing Watch (GFW) will award prizes totaling $100,000 USD, available to all individuals, companies, or organizations regardless of geographic location, for submissions on the Open Source track”