Reusing Convolutional Activations from Frame to Frame to Speed up Training and Inference

Re-assign community