Derp Learning is a Python package that collects data, trains models, and then controls an RC car for track racing.
Getting Started
Software
The developers of this project primarily use Ubuntu 16.04 on both x86_64 and aarch64 architectures.
Inside the install folder there are series of prerequisite install scripts. Run prerequisites.sh to run them. Then setup.py script will install the Python package.
cd install
bash install.sh
Hardware
You will need access to a drill to cut holes in the car's chassis and into the PVC sheet or 3D printed camera pylon. Select one (or more) of the compute options. TODO detailed instructions
All of the following commands need to be run from the src folder.
Collect Data
On the car run:
python3 drive.py
Data by default is saved to files in the folder /data/ which is created in the parent directory of /derp_learning/
The data can be moved by swapping the SD card if the derplearning directory is located there or by using ssl rsync from the directory you want to move the data to on your device:
To move label and train a model on collected data use the shell script pipeline.sh. This is the ideal way to deploy a model trained on same day collected data. This option may be used instead of manualy performing the below steps.
Once a model is deployed to the car it can be loaded by pressing the appropriate button and given control of the vehicle by pressing the playstation button.