Frederick Tung
Papers
8
Total Citations
404
H-Index
7
About
Frederick Tung is a researcher whose work bridges deep learning, computer vision, and autonomous robotics. His primary research areas include neural network compression, camera relocalization, and mobile robot navigation in challenging environments. Tung’s most impactful contribution is his work on deep neural network compression, where he pioneered an in-parallel pruning-quantization method that simultaneously reduces model size and computational cost while maintaining accuracy—a critical advancement for deploying deep networks on resource-constrained devices. This paper has garnered 138 citations, reflecting its significance in the field. In robotics, Tung has made notable strides in autonomous navigation, developing systems capable of operating safely in uneven and unstructured indoor environments (102 citations). He has also advanced camera relocalization through innovative regression forest techniques that exploit both points and lines for accurate pose estimation in RGB-D systems, with applications in self-driving cars and virtual reality. His work on the Raincouver Scene Parsing Benchmark addresses the critical challenge of self-driving perception in adverse weather and nighttime conditions, providing a standardized evaluation framework that has become a valuable resource for the community.
Research Focus
Key Achievements
Top Papers
- 1Deep Neural Network Compression by In-Parallel Pruning-Quantization138 citations · 2018
- 2
- 3Backtracking regression forests for accurate camera relocalization60 citations · 2017
- 4
- 5
- 6Backtracking Regression Forests for Accurate Camera Relocalization14 citations · 2017
- 7
- 8