Network Training

Copy the training data onto your laptop.

Store the training data in the following folder: RVSS2019-WS/on_laptop/dev_data/

You can copy the training data by executing the following command from the dev_data folder on your laptop:

scp -r pi@192.168.50.5:~/RVSS2019-WS/on_robot/collect_data/data/* .

Train your network.

We have not provided an example training script for you, you will need to create this on your own! We recommend looking at pytorch tutorials for training classification or regression networks (depending on what you have chosen when you designed your network). You can find some example tutorials saved as pdf's on the documentation page of this website.

We have created a custom dataset class for you, to make it easier to load in the images and use them with the pytorch DataLoader. This custom dataset class, along with a test() function that shows how to use it, is stored in RVSS2019-WS/on_laptop/steer_net/steerDS.py

Once your network has finished training, make sure to save the network weights! For a network named steerNet, you can use the following command to do this:

torch.save(steerNet.state_dict(), "steerNet.pt")

Now you can finally deploy your network on the robot and test the performance!