Applying Transfer Learning on Inception V3 model (weights trained on Imagenet) for the Oxford TV Human Interactions dataset. The network gets as inputs images extracted every 5 frames from videos.
