About me
Currently, I am a Senior Research Engineer at Tencent working with the Media Lab led by Shan Liu. My role involves research and development of 3D Graphics coding technologies for meshes and point clouds. I am also actively involved in standardization activities for 3D data in collaboration with other companies. I earned my Ph.D. from University of Southern California where I was advised by C.-C. Jay Kuo. My dissertation was related to 3D perception from point cloud data using unsupervised machine learning techniques. Previously, I interned at Sony R&D Center USA and InterDigital Video Lab where I contributed to AI-based 3D point cloud compression exploration activities in MPEG. I also keep an interest in Computer Vision, Machine Learning and exploring emerging technologies in general.
Doctor of Philosophy (PhD), Electrical Engineering, 2023
University of Southern California, Los Angeles, CA
Master of Science (Honors) in Electrical Engineering, 2020
University of Southern California, Los Angeles, CA
Bachelor of Engineering, Electronics and Telecommunication, 2018
Savitribai Phule Pune University, Pune, India
Experience
- Research and development of ongoing AOMedia VVM standard on Static Polygonal Mesh Coding
Tencent Media Lab Supervisors - Xiaozhong Xu, Shan Liu
- Researched methods to improve compression of static meshes by exploiting mesh symmetry
Sony R&D Center US Laboratory Supervisors - Danillo Graziosi, Alexandre Zaghetto, Ali Tabatabai
- Developed a deep predictor network for inter-prediction in dynamic dense point cloud compression
- Designed rate control mechanism in deep learning based point cloud compression methods using gain/inverse gain units
- Proposed unified neural network architecture and joint training approach for I- and P-frame compression
- Achieved BD-Rate of -10% over SOTA deep learning method with fewer parameters and BD-Rate of -60% over V-PCC
InterDigital Video Lab Supervisors - Dong Tian, Jiahao Pang
- Designed intra-/inter-mode decision module for dynamic point cloud compression
- Proposed training of scene flow estimation methods with unsupervised RD loss for dynamic point cloud compression
- Improved performance of dynamic LiDAR compression over G-PCC using deep learning techniques
Certifications
Publications
PDF Code Dataset Poster Slides Video Source Document Custom Link
Contact
- hello@pranav3D.com
- +1-747-334-8822
- Sunnyvale, California, USA