Did you try to run the python command ? Did you see that there is an issue on which others requested to improve the Readme file ? If by experiments and by looking the source .py code you are not able to identify what to do, I would recommend you add your request at the Issue tab on the Git page, as 2 others have done.
That's right, it just a dataset , you need a framework for Machine Learning like lite tensorflow on the Pi but the learning time will slowd due to CPU core and dont have GPU support.
You won't be learning on a pi but inference. You train on a PC with gpu and deploy the model to the pi. Coral or gpu on the pi maybe enough to perform inference and it is all depend on your model size and precision.
You won't be learning on a pi but inference. You train on a PC with gpu and deploy the model to the pi. Coral or gpu on the pi maybe enough to perform inference and it is all depend on your model size and precision.