Mirror-Aware Neural Humans (3DV 2024)
Learning human body pose, shape, and appearance from mirror images and 2D detections.
I am a Ph.D student in Computer Science at The University of British Columbia (UBC) working with Prof [Kwang Moo Yi, Helge Rhodin, and Andrea Tagliasacchi] with interest on generative models, neural human avatars and 3D reconstruction.
I was previously a Google and Facebook scholar at the African Masters in Machine Intelligence program (AMMI). I am blessed to have worked with Dr. Gaurav Bharaj and Dr. Pablo Garrido at Flawless AI, mentored by Dr. Gamaleldin Elsayed from Google Brain, and taught by Prof. Leonid Sigal at UBC.
My work has largely focused on solving problems in human motion learning, 3D reconstruction, self-supervised learning, and generative body models.
Learning human body pose, shape, and appearance from mirror images and 2D detections.
Understanding Needs for University Curricula.
Estimating the metric distance between persons from a single video.
An End-End integral pose regression for 2D and 3D Human pose estimation.
Instance segmentation was carried out on a nuts custom dataset that was trained on an object detection library (Detectron2).
A simple object tracking system that detects objects on all video frames using predictions from a detectron model and links these detections across the temporal domain.
Estimating the location of 68 distinguishing keypoints on human faces.
Automatically generating captions that describe the content in images.
Implementing a real-time object detection with Tensorflow and OpenCV.
Implementing SLAM (Simultaneous Localization and Mapping) for a 2 dimensional world!.