Closed crypdick closed 7 years ago
The first image is the default emulator window size. The second is what Nvidia used in their research paper. I used their model and I didn't want to change the network size to make things easier. That is why you see a difference.
It shouldn't make a difference as the data is all still there albeit resized or skewed a bit. It should be possible to change the network to work on the same aspect ratio it just involves a bit of math. Should be easier as well now that we're using Keras.
Makes sense. Is that why the output layer has 10 units when there are only ~5 controls?
Nah thats because originally I was going to use more of the controller than I ended up using.
I noticed that the Sample class has shape (200, 66, 3) while the raw images have an aspect of (640, 480, 3). Why is there a difference in aspect ratio?