Path finding autonomous 1/10th car

Hello, I wanted to learn how make a small autonomous vehicle navigate its environment using a deep neural network and a camera. Here is a GitHub project that I created recently from the result of such desires.

https://github.com/bastulli/AutoCarJetsonNano

So far its able to record its inputs from a PlayStation controller over Bluetooth and does not a WiFi connection to switch from training to autopilot. You can offload the images from the camera with logged input data to a USB and train a neural network in Jupyter Notebook. I am happy of what I learned with this project and hope that others can utilize this or parts for other robotic projects.

hi! Thanks for sharing your code, I have some experience with keras but your train and inference scripts seem very approachable. I definitely will check and test pytorch.
I have a question regarding the hardware part, how did you power the jetson nano? I saw a lipo battery but this battery is rated ad 7.4 volts. Thanks!

I used this https://www.amazon.com/gp/product/B07H6LB4J4/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1. It’s typically used for chagrining a cellphones but its enough to power the board at half power through USB. I hope to switch to a bigger battery soon so I can fully utilize the Jetson Nano. The 7.4 volts lipo is to power the electronic speed controller that drives the motor on the car.