scollins83 / carnd-behavioral-cloning-p3 Goto Github PK

• Model contains dropout layers

• Testing by running through simulator and validating the car stays on the track for one lap.

Create a version of the current code with a simple augmentation generator made through base Keras functionality.

Upload zip file of training data from local machine to AWS

• Convolutional NN with 3x3 filter sizes and depths between 32 and 128 (model.py lines 18-24)

• Uses ReLU Layers

• Uses a Keras Lambda layer

Currently, the cv2.imread() function at line 84 in function 'get_images_and_measurements' is returning a 'None' type with a supplied path, even when applied on it's own without the rest of the function. This is causing the shape of the tensor going into the model to error out.

• Model used ADAM optimizer, so learning rate is not tuned manually