DingoCar: Perception module Small scale, self-driving car
summary
Small-scale self-driving kit runs TensorFlow on Raspberry Pi (Jetson Nano support) with a camera, IMU, PWM motor outputs and LiPo power; data collection records ~5,000 frames of steering/throttle into a donkeycar model and produces a ~3 MB trained net. Training on Google Colab drops from ~15–20 minutes locally to about a minute on GPU. The model drives steering-only reliably on track but still fails when reflections or irrelevant objects dominate the image — saliency maps expose those failure modes.
AI-written summary · about
chapters (7)
related talks
33mSuper Donkey Cars
01h 25mHow to Build Your Own Self Driving Toy Car
35mGetting started with a self-driving RC car
31mAutonomous driving, how Python drives Formula Racing cars
47mUnlocking Data-Driven Apps with TensorFlow
23mRacing Robocars
28mMachine Learning on the Edge
31mDeep learning for self-driving vehicle technology