Geng Yang

Geng is a Ph.D student in the School of Communication Engineering at Xidian University in China, supervised with Prof. Yunsong Li and Prof. Jie Lei. She expects to graduate in Dec. 2024.

Her research focuses on algorithm-hardware co-design for DNN architecture and high-performance FPGA-based hardware acceleration. Over the past three years, she has been dedicated to the development of an end-to-end DNN model inference acceleration system that integrates hardware-friendly model pruning and quantization, and FPGA-based hardware architecture. She has actively participated in one National Natural Science Foundation of China and two large national projects under related topics. Geng’s work has been published in high-impact journals such as TRETS, TGRS, JSTAR, etc. In recent months, she has been collaborating with Professor Zhenman Fang from Simon Fraser University on the hardware acceleration of larger and more complex DNN-Stable Diffsion inference.

Geng's Email: gengyang [at] stu.xidian.edu.cn

Email / CV / Google Scholar

Research Interests

FPGA-based Deep Learning Inference Acceleration
Algorithm-Hardware Co-design

Selected Publications

Full publication list

	HyBNN: Quantifying and Optimizing Hardware Efficiency of Binary Neural Networks G. Yang, J. Lei, Z. Fang, Y. Li, J. Wang, W. Xie [TRETS 2023] [FPT 2023 Journal Track] Paper/ FPT 2023 Oral report
	Guided Hybrid Quantization for Object Detection in Remote Sensing Imagery via One-to-one Self-teaching J. Zhang, J. Lei, W. Xie, Y. Li, G. Yang, X. Jia [TGRS 2023] Paper
	Algorithm/hardware Codesign for Real-time On-satellite CNN-based Ship Detection in SAR Imagery G. Yang, J. Lei, W. Xie, Z. Fang, Y. Li, J. Wang, X. Zhang [TGRS 2022] Paper/ Code
	A Low-complexity Hyperspectral Anomaly Detection Algorithm and its FPGA Implementation J. Lei, G. Yang, W. Xie, Y. Li, X. Jia [JSTAR 2022] Paper / Code / Excellent Oral report in the 6th National Imaging Spectroscopic Earth Observation Symposium
	Rank-Aware Generative Adversarial Network for Hyperspectral Band Selection X. Zhang, W. Xie, Y. Li, J. Lei, Q, Du, G. Yang [TGRS 2022] Paper

Conference Abstracts

E4SA: An Ultra-Efficient Systolic Array Architecture for 4-Bit Convolutional Neural Networks, G. Yang, J. Lei, Z. Fang, J. Zhang, J. R. Zhang, W. Xie, Y. Li (FPGA 2024 poster)
HyBNN: Quantifying and Optimizing Hardware Efficiency of Binary Neural Networks, G. Yang, J. Lei, Z. Fang, Y. Li, J. Wang, W. Xie (FCCM 2023 poster)

Patents

Object Detection Joint Pruning and Quantization for SAR Imagery, J. Lei, J. Wang, G. Yang, W. Xie, Y. Li. (CN202111488427.7 on December 8, 2021)

FPGA-based Hyperspectral Anomaly Detection System, 2021, J. Lei, G. Yang, M. Zhang, W. Xie, Y. Li, T. Jiang, K. Liu, L. Gao. (CN202110484719.7 on April 30, 2021)

Contest Award

	Our team supervised by Prof. Jie Lie won the second prize in the national system contest and received ¥3,690,000 project funding. This contest focuses on DNN model inference on a cloud FPGA system. Geng Yang is student leader.
	Our team supervised by Prof. Weiying Xie won the third prize in the national competition and received ¥950,000 project funding. This contest focuses on DNN model inference on an embedded FPGA system. Geng Yang is student leader.

Honors and Awards

Innovation Fund of Xidian University in 2022,2023
GuoRui Scholarship of the 14th Research Institute of China Electronics Technology Group Corporation in 2021,2022
Second Scholarship of Xidian University in 2020,2021,2022,2023
Collaboration-Innovation Scholarship of China Electronics Technology Group Corporation and Xidian University in 2020
Outstanding Graduate Student of Xidian University in 2019

Professional Services

Conference Secondary Reviewer
- FPGA'24
- DAC'23, 24
- DATE'24

FPGA Acceleration Demo

CNN-based Ship Detection in SAR imagery

This is a demo presentation for Paper:Algorithm/hardware codesign for real-time on-satellite CNN-based ship detection in SAR imagery.

Thanks to the laboratory partners who worked together to finish this topic!

Last updated on 12/2023 EDT.
Webpage template