Pre-trained model issue and training model input data issue

llSourcell / How_to_simulate_a_self_driving_car

This is the code for "How to Simulate a Self-Driving Car" by Siraj Raval on Youtube

430 stars 306 forks source link

Pre-trained model issue and training model input data issue #13

Open duzumich opened 7 years ago

duzumich commented 7 years ago

Pre-trained model issue : I ran the pre-trained model, the car always ran out of the track, does anyone meet the same problem?

training model input data issue: I built a folder called 'data', and put the images and driver_log file in the folder, but it doesn't work. Was it wrong? Could you give me some advice? Thanks

duzumich commented 7 years ago

I have solved my problem, it is because the input image sizes issues in the "random_shadow" function. If met with the same issue, you may change with "x1, y1 = image.shape[1] np.random.rand(), 0 x2, y2 = image.shape[1] np.random.rand(), image.shape[0] xm, ym = np.mgrid[0:image.shape[0], 0:image.shape[1]]" to solve your problem.