Research

Deep learning for computer vision applications

  • High accurate object recognition (unconstrained face recognition)

    • ResNet 101 with center loss

    • Fast 3D face model approximation using a deep learning pipeline

  • Object detection and tracking

    • RCNN, Faster RCNN, MDNet

    • Mask RCNN

    • GANs to generate infra-red object images

  • Motion prediction from Mocap data

  • high confidence detection, tracking, and classification

    • computer graphics based data augmentation

    • image/video translation for generating training data

    • loss functions

    • multiple objects tracking and motion pattern analysis

  • 3d inference

    • classical optimization framework

    • learning 3d representations

    • inferring 3d primitives from 3D data

    • deformable objects