Yellow & Blue is a mini-game that allows the player to "play around" with ML-Agents, by switching between exported graphs and lessons within the game. They can also play against trained AI in order to grasp how hard the task would be for them, how bad the training went, etc.
The environment consists of two circles, Blue and Yellow. Blue must reach Yellow, who implements basic steering behavior (flee + evade). More information on the training environment can be found on the Github repository.
The game interface allows user to view and interact with 32 agents. They can:
Control any moving object
Change exported graphs ("checkpoints")
Change lesson reset parameters ("lessons")
For now, Windows Build only is available on the game page, but other build platforms (MacOS, Linux, WebGL(????)) will soon be supported.
Sound design has been made by IaMed.
The source code can be found on the Github repository.
Thanks to the ML-agents developers and Unity for introducing cutting edge technology to non-savvy devs. <3