Knowledge distillation: Train a Student model from a Teacher, retain most of its performance
Estimating 3D human pose from images
Data association over graphs, network consistency.
Single and multi-camera data association: tracking and person re-identification.
Analysis of confocal microscopy based plant live-cell imaging data.
Wide-area video surveillance: human activity recognition, forecasting and anomaly detection.
3D segmentation of neuro-images, semantic segmentation.
We seek solutions to computer vision and machine learning problems arising in numerous application areas involving visual data analytics. These include knowledge distillation, human pose estimation, person re-identification in camera networks, super-resolution imaging, domain adaptation etc. to name a few. We are also keen to explore how visual analytics can be utilized in answering some of the most fundamental questions in biology and healthcare.
