Home - GarrettHaley/SelfDrivingRCCar GitHub Wiki
Welcome to the Self Driving RC Car Wiki!
OVERVIEW
To build a self driving rc car with minimal functionality, using novel deep learning techniques and electronics. The system will involve a raspberry pi, a movidius NCS, a pi camera connected to a arduino controlled DIY RC car. The car will recognize signs and then take actions depending on the sign it comes across.
GOALS
Create a dataset of stop sign images, with over 1000 images for training. To do this we could do it manually or utilize a python web scraper. Or this dataset https://www.kaggle.com/c/traffic-sign-recognition/data Then we will need to process the images to 320x240, and apply rotation/blurring and various other affects. Split data into Training/Validation and Testing Sets. Train a CNN using keras on the Dataset. Design a robust but basic architecture that handles for overfitting and underfitting using various proven techniques. The goal here should be to maximize training/testing accuracy and to minimize training/testing loss. Get NCS working on the raspberry pi Upload the Trained network onto a Neural Compute Stick, and interface with raspberry pi. Configure a RC car to work with a raspberry pi. Need to figure out how to control the car using the raspberry pi’s gpio pins. Write a program that utilizes the trained model and steers the car. Test the program
SPECIFICATIONS/TOOLS
- Raspberry Pi 3
- Movidius Compute Stick
- RC car
- Arduino
- Wires
- Raspberry Pi camera
- Stop Sign
- Battery Pack
- SD Card
- Blocks
- Arduino firmware
- RC car mechanical design
- Raspberry Pi Neural Compute Stick
- Node.js interface on raspberry pi
- Dataset on the cgrb server in share folder
- CNN architecture
- Training the architecture
References and Sources
- https://medium.com/deep-learning-turkey/a-brief-guide-to-intel-movidius-neural-compute-stick-with-raspberry-pi-3-f60bf7683d40
- https://makezine.com/projects/build-autonomous-rc-car-raspberry-pi/
- https://www.kaggle.com/c/traffic-sign-recognition/data
IRC Channel:
https://webchat.freenode.net/ Channel Name: OSUMLC