Machine Learning

Optimization and sampling algorithms
Continuous time series models, latent variable models and structured prediction
Deep learning, large scale supervised and semi-supervised methods, kernel methods

Computer Vision

Human motion sensing (2D localization, 3D reconstruction, action recognition) in video with emphasis on the cognitively relevant monocular-limit
Visual recognition. Image and semantic segmentation in RGB and RGB-D data
Computational modeling of eye movements from large-scale recordings

Large-Scale Datasets, Evaluation Benchmarks and Software

Large-Scale Datasets, Evaluation Benchmarks and Software VOC Actions: image human eye movement recordings for PASCAL VOC Actions from a Single Image Dataset (1 million human fixations, 86 hours of cummulated exposure time, multiple task conditions).
Actions in the Eye: video human eye movement recordings for Hollywood-2 and UCF sports (669.187 human fixations, 92 subject-video hours, multiple task conditions).
Percept3D: a dataset for analysis of human pose re-enactement (images, 3d motion captures, eye movements).
Human3.6M: 3.6 million human pose dataset and software (multiple viewpoint 2D data, 2D and 3D motion capture and time of flight data, as well as human body part labeling).