Atsuto Maki, Professor

Research Interests: Machine Learning, Computer Vision, Interaction

Deep ConvNet Representations:

Network Regularization - Feature Contraction

(demo code by Vladimir Li)

Factors of Transferability

A baseline for visual instance retrieval

Motion and Object Recognition:

Correlated Probabilistic Trajectories

A new distance for scale-invariant 3D shape recognition and registration

Demisting the Hough transform

Geotensity, Matching Measures:

Geotensity constraint: combining motion and lighting

SCVD: Sum of conditional variance of differences

Hyper-Patches for 3D Model Acquisition and Tracking

Learning-based stereo

Industrial Applications:

Subspace methods

Lighting environment estimation

Computational cameras

Clustering Techniques:

Random Forest Clustering

MCL-Superpixels: Compact Markov Cluster Algorithm for generating homogeneous superpixels

Attention, Biological & Machine Vision:

Disparity Selection in Binocular Pursuit (Video)

Stereo Vision in Attentive Scene Analysis