From the moment we introduced it last September, the developer community has wholeheartedly embraced the ML-Agents Toolkit beta.  We’ve seen people use this toolkit to create environments, train agents for different use cases, and even apply them to their games and apps.

Small project I’ve done to experiment deep learning #unity3d ML-Agents. Very promising. Congrats to @awjuliani and his team! pic.twitter.com/GS0E4CwTu0

— Maxime Plantady (@MaximePlantady) October 2, 2017

I’ve been playing with machine learning and Unity a lot lately. This was one of my first tests. The blue things are trained to push a block to jump over the wall. Left 10k iterations, middle 100k, right 500k. pic.twitter.com/0mMRvRGJqy

— Renaud Forestié (@Reuno) December 18, 2017

All of this has inspired us to find new ways to connect our developers for the exchange of exciting training ideas, to inspire new use cases, and to reward community contributions. On Dec 8, 2017, we launched a dedicated discussion channel on Unity Connect and kicked off our very first ML-Agents challenge. Today, we are announcing the competition’s winners, determined by voting from the entire Unity Machine Learning team.

And the winners are…

No.1 Pass the butter // Pancake bot

Making the perfect pancake doesn’t have to be hard, nor does it require any real cooking skills.  Plus, who doesn’t like pancakes?  Or robots for that matter?

No.2 Metal Warfare – Real Time Strategy game

Excellent take on how machine learning can be used in strategy games to train and evolve AI.

No.3 Hide/Escape – Avoidance of Pursuing Enemies

Based on the stealth genre of games, a new way to hide from traditional game AIs.

For this challenge, we also have a special award for the project which received most likes from the community.

Most Liked Project Award

Satellite AI

2nd Most Liked Project Award

Jump(跳一跳) with ML-Agents

In addition to the four prize winners above, we also selected the following four projects as the ML-Agents Creative Application Award winners.

Autonomous Cars

SpaceY

Vehicle Environment Static Environment

Fun Waves…

Congratulations to the winners and thanks to all who participated in our very first machine learning challenge! All winners will soon receive instructions on how to claim their prizes. And we will feature all winning projects on the Unity3d.ai official portal.

Meet us at GDC 2018

We will release ML-Agents Toolkit v0.3 during the week of GDC 2018. And we’d like to invite you to meet with our team to learn more about the new features! We have a series of Machine Learning programs at this GDC, which you can see below: and here is a list for you to preview.

At GDC West Hall:
Unity for Deep Learning: ML-Agents Toolkit explained

At Unity Central:

1. Unity Machine Learning: ML-Agents Toolkit 0.3
2. Unity ML-Agents Toolkit Hands-on Lab
3. Unity Machine Learning office hours

Unity Machine Learning Roundtable

If you have used ML-Agents or machine learning in games, we want to hear your experiences, insights, and feedback. Fill out the sign-up form to be part of the future of Unity Machine Learning development. Sessions will be held March 21, 11:00 – 12:00 and March 22, 2:30 – 3:30.

Looking forward to meeting you at GDC 2018!