Question How to feed scene data to a model I am training using ML-Agents in Unity

I want to train a model using the ML-Agents package. In the training process, I would like to capture a sequence of screenshots of a scene while it is in play mode, and feed those snapshots to the model I am training in real-time (as the input data). Is there any way I can do this in Unity? Like, a script that starts screenshotting the scene frame by frame and passing it to the model in real time? Are there any packages that can help me achieve this?

 

Is your goal to have the data in the scene saved ? Or the positions/elements presents visually displayed in an image fed ?