for collecting training data run sudo python complete_control.py
Training the network
training the CNN with the gathered data to predict steering angle and acceleration
Predictions and further visualizations
after successfully predicting steering angle and acceleration I decided to write the predictions directly on the image
further implemented an arrow that visualizes the direction the network predicted to steer. The length of the arrow as well symbolizes the intensity of acceleration.
Updates and Changes:
trained a smaller model for predictions because loading and calc with the previous model was too calculation costly. Therefore I trained another - smaller model- and increased the size from 331 MB to 1.4 MB. While the accuracy still is up to 92% compared to 97% for the bigger model! huge success in my eyes!
