Blazing Fast Self Driving RC Car
200FPS Image Analysis on Raspberry Pi
Full Convolutional Neural Networks are simply too much processing for a Raspberry Pi 3B - especially at 200FPS.
A more custom approach is needed running on a constrained environment.
- Support speeds of at least 30mph
- Adapt to any visual conditions:
- Indoor support with glare from indoor lighting
- Outdoor support with glare from sun and shade over the course
- Learn extreme driving modes like sliding turns and drifting
- Map course solely from visual input
- Pixel Online Learning Module
- Learn Pixel Colors Online (during runtime) to adjust to lighting conditions
- To ensure stability, use only verified results - strategically using a more expensive algorithm like edge detection to confirm learning results
- ./buildme
- ./run