Multi-Agent Behavior Challenge | CVPR 2021

By  vrv

๐Ÿ’ก Our friends at Northwestern & Caltech are working to solve an important problem in neuroscience.

How to understand the structure of behavior, its control by the brain, and its dependence on context, experience, and more? ๐Ÿง 

But, the researchers who are studying behavior (our friends) actually end up spending countless hours just painstakingly annotating videos of animals and their actions - losing important time that they can dedicate to doing more research. ๐Ÿ˜”

Which led us to host an interesting...and lucrative challenge on AIcrowd! โœจ

Can we use machine learning to defeat this 'Annotation bottleneck' for researchers?

๐ŸŽฏ There are 3 goals for this challenge

  1. To develop methods to classify human-defined behaviors from tracked pose trajectories in a large dataset of videos of socially interacting mice
  2. To fine-tune classifications to different annotator styles
  3. To learn to recognize new behaviors of interest from limited training examples.

๐Ÿ’ฐ These tasks have a total prize pool of $9,000!

Task 1

Classical Classification


Task 2

Style Transfer Task


Task 3

Learning New Behavior



๐Ÿคฏ On top of this, we also have total $10,000 Amazon SageMaker credits for first 50 participants/teams to beat the baseline scores by 5%!

And..not only that, eligible winners will also be invited to speak at the Multi-Agent Behaviour Workshop at CVPR2021! ๐ŸŽค

Check out the competition here

Written by



You must login before you can post a comment.

You may also like...