MABe Challenge - Social Action Recognition in Mice

Overview

This competition challenges you to build machine learning models that can recognize over 30 different social and non-social behaviors in mice based on their movements captured in video recordings. Your goal is to accurately identify these behaviors, similar to a trained human observer, while accounting for variations in data collection equipment and motion capture techniques. The work done here will help scientists automate behavior analysis and gain deeper insights into animal social structures.

Requirements

Participants must be at least 18 years old to compete. The competition is open to everyone globally. You'll need to submit your solution as a Kaggle Notebook, which has run-time limits of 9 hours for both CPU and GPU. Internet access must be disabled for submissions, and you can use freely available external data and pre-trained models. Your submission file must be named submission.csv and follow a specific format that includes row ID, video ID, agent ID, target ID, action, start frame, and stop frame for each identified action.

Prizes

The top five participants will receive cash prizes. First place wins $20,000, second place gets $10,000, and third place is awarded $8,000. The fourth prize is $7,000, and the fifth prize is $5,000. In total, there is $50,000 in prize money available.