Road Signs

Traffic Sign Detection

For the traffic sign detection we used:

• Cascade Classifier for traffic sign region detection.

• SVM classifier for classification of the traffic sign type.

To train the models, we created a custom dataset with images from the real track (taken from videos given by the organization), images from the simulator, and data augmentation (e.g. overlapping of traffic signs on images of the road).

Semaphores

The semaphores transmit their state through UDP messages at all times. Knowing that, we were able to simply listen to the messages and take the state of the semaphore nearest to the car. Then, depending on the state of the semaphore, the car would stop when it reached the horizontal line in the road (that indicates where it should stop), and resume its course once it turned green.