research Learning Pixel Trajectories with Multiscale Contrastive Random Walks Self-supervised learning of optical flow, keypoint tracking, and video object segmentation via multiscale contrastive random walks on pixel-level space-time graphs. Motion Estimation and Inverse Problems in Tagged MRI Physics-constrained deep learning methods for motion estimation, tracking, and blind inverse problems in tagged MRI, with applications to tongue and cardiac imaging. Aorta Registration and Deformation Analysis CT-based deformable image registration methods for measuring and characterizing thoracic aortic aneurysm growth, with clinical validation. Weakly-Supervised Vitiligo Segmentation A weakly supervised framework for segmenting vitiligo lesions in skin images using saliency propagation on superpixel-based graphs. Highlight Removal in Facial Images A method for removing specular highlights from facial images while preserving natural skin appearance. 6D Pose Estimation from Single RGB Image Estimating 6D object pose from a single RGB image using a backpropagatable PnP formulation. Video Synopsis for Surveillance Efficient keyframe extraction and dynamic video synopsis methods for condensing long surveillance videos. fun Q-value Based RL on Atari Games Implementing and comparing deep reinforcement learning algorithms (DQN, Double DQN, Dueling DQN) on Atari game environments.
