Researchers have developed an add-on module that helps artificial intelligence systems called convolutional neural networks, or CNNs, to fill in the gaps between video frames to greatly improve the network's activity recognition.