We provide some highlights of the topics we work on, including who works on them and the latest publications on the topic.
Deep Learning is rapidly becoming a game for the powerful: if you don’t have huge amounts of data, you cannot compete. Inductive priors (or biases) are choices, built into a neural network, that make the network need less data to learn a good function of the dataset. That is, if you choose the right prior!
A video is not merely a collection of static independent images. Video includes motion, action, interaction, dynamics, causal effects, long term behavior, emotion. Our research includes investigating how to (deep) learn motion representations, exploit dynamics, classify actions, localize actions in space and time, automatic emotion recognition, pose estimation and object tracking.