Self-driving car using Convolutional Neural Network. The project was done with the Metrowest Boston Developers Machine Learning Group.
Refer to cnn_auto_driving_car_project.pdf for details about the project.
The project utilizes TensorFlow and Kiras Python libraries. A good reference is "Hands-On Machine learning with Scikit-Learn & TensorFlow" (Geron) from O'Reilly. This includes steps to set up a virtualenv using PIP and full instructions and examples on Tensorflow and Scikit.
This git repository is available via: git clone https://github.com/cwinsor/metrowest_scikit_tensorflow_cnn_car.git
The repository includes the following documentation, data, and Python source:
cnn_auto_driving_car_project.pdf - documentation about the project model\Learn To Drive.ipynb - jupyter notebook which trains the CNN model\weights.hdf5 - weights file for the trained CNN rasp_pi\self_driving_car_capture.py - file run on the Raspberry Pi to capture image and steering data rasp_pi\self_driving_car_drive.py - file run on the Raspberry Pi to drive the car using the trained CNN rasp_pi\self_driving_car_test.py - file used during development to verify GPIO and camera functionality. lib\ - user and architect libraries and examples as follows: lib\test*.ipynb - jupyter notebooks with examples demonstrating use of libraries lib\metrowestcar_dataset.py - user library to read dataset lib\metrowestcar_display.py - user library to display annotated images lib\metrowestcar_dataset_architect.py - architect library to build dataset from raw data files lib\metrowestcar_file_io.py - architect library to read raw data files data\ directory where the dataset is kept data_raw\ directory where raw flatfile data is kept