Here's a video about this project
- Bash on ubuntu on windows
- Python 2.7
- Numpy
- SciPy
- Tensorflow
- Keras 1.1.0
- udacity-driving-reader from rwightman, used to extract images from ROS bag files
the dataset can be downloaded from here:
training: link CH2_002
testing:link : "scenario1" folder then go to "data" folder then finally video data set is on "car" folder, [scenario1 -> data -> car -> car1_camera1.mp4 ]
Directory structure for data:
- data
-- models
-- test
--- center
---- 1477431483393873340.jpg
-- train
--- center
---- 1477431802438024821.jpg
Change data_path
value in config.py
to point to this data directory.
To pre-process training data, run:
python train_data.py
To pre-process test data, run:
python test_data.py
These pre-processing scripts convert image sets to numpy arrays.
To train different models, run:
python model.py
You can change these parameters in the config.py
file:
To predict steering angles from test data, run:
python drive.py
- Visualizing predicted steering angles
To visualize model predictions on test data, run:
python renderview.py