Overview / Usage

The project is using RL Coach and Open VINO toolkit.
Phase 1:
Using RL Coach we are traing the experiments such as Cartpole,StarCraft 2 simulation when we are doing the simulation we are dumping GIF files of training after a required interval.Now from this GIF files we segragate images using a python script.
This python script creates lots of images of the experiment ,we break the images down to the most recent ones too.
After that we created an image dataset from which we trained and created customized Mobile SSD custom datset using Intel Optimized Tensorflow.
Phase 2:
Using Intel Open VINO toolkit we optimize the customized dataset creating and xml and bin file
Now using the Inference engine we are able to distinguish between the earlier stages of training as well the last one performed
finalized for Intel Architecture.

Methodology / Approach

We have used RL Coach inside the Ubuntu 14.04 OS.We trained different environments such as Cartpole,Star Craft 2 and others.
Now using Intel Optimized python we wrote scripts to seggrate the GIF dumps to images.
After that the image dataset obtained is customized using Intel Optimized Tensorflow to create a custom mobile ssd dataset.
We then use Intel Open VINO toolkit to get the optimization done.
Using Inference engine we are able to distingush which is the last stable simulation for the RL Coach training done and which one is the earliest.

Technologies Used

RL Coach
Intel Optimized Python
Intel Optimized Tensorflow
Intel Open VINO toolkit.

Collaborators