Yixiong Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Yixiong Yang has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Huazhong Yang
Tsinghua University
6 shared publications - Yongpan Liu
Tsinghua University
6 shared publications - Jinshan Yue
Institute of Microelectronics of the Chinese Academy of Sciences, Beijing, China
3 shared publications - Fang Su
Department of Electronic Engineering, Tsinghua University, Beijing 100084, China
2 shared publications - Jingyu Wang
Department of Electronic Engineering, Tsinghua University, Beijing, China
2 shared publications Xiaoyu Feng
Department of Electronic Engineering, Tsinghua University, Beijing, China
2 shared publications
Latest Publications
- Open Access
- 4 December 2024
- Published by Qeios Ltd in Qeios Preprints
MLI-NeRF: Multi-Light Intrinsic-Aware Neural Radiance Fields
Current methods for extracting intrinsic image components, such as reflectance and shading, primarily rely on statistical priors. These methods focus mainly on simple synthetic scenes and isolated objects and struggle to perform well on challenging real-world data. To address this issue, we propose MLI-NeRF, which integrates Multiple Light information in Intrinsic-aware Neural Radiance Fields. By leveraging scene information provided by different light source positions complementing the multi-view information, we generate pseudo-label images for reflectance and shading to guide intrinsic image decomposition without the need for ground truth data. Our method introduces straightforward supervision for intrinsic component separation and ensures robustness across diverse scene types. We validate our approach on both synthetic and real-world datasets, outperforming existing state-of-the-art methods. Additionally, we demonstrate its applicability to various image editing tasks. The code and data are publicly available at https://github.com/liulisixin/MLI-NeRF.
- 4 December 2024
- Published by Qeios Ltd in Qeios Preprints
Open AccessMLI-NeRF: Multi-Light Intrinsic-Aware Neural Radiance Fields
Current methods for extracting intrinsic image components, such as reflectance and shading, primarily rely on statistical priors. These methods focus mainly on simple synthetic scenes and isolated objects and struggle to perform well on challenging real-world data. To address this issue, we propose MLI-NeRF, which integrates Multiple Light information in Intrinsic-aware Neural Radiance Fields. By leveraging scene information provided by different light source positions complementing the multi-view information, we generate pseudo-label images for reflectance and shading to guide intrinsic image decomposition without the need for ground truth data. Our method introduces straightforward supervision for intrinsic component separation and ensures robustness across diverse scene types. We validate our approach on both synthetic and real-world datasets, outperforming existing state-of-the-art methods. Additionally, we demonstrate its applicability to various image editing tasks. The code and data are publicly available at https://github.com/liulisixin/MLI-NeRF.
- 11 June 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-2
A Demonstration Platform for Large-Scaled Point Cloud Network Based on 28nm 2D/3D Unified Sparse Convolution Accelerator
3D point cloud processing plays an important role in many emerging applications such as autonomous driving, visual navigation, and virtual reality. It calls for hardware acceleration of multiple key operations, including 3D Submanifold SCONV, 3D non-Submanifold SCONV, and 2D SCONV. This work presents a 2D/3D unified sparse convolution accelerator for large-scale voxel-based point cloud networks. The chip is fabricated in TSMC 28nm CMOS technology to achieve 3.3-16.9 FPS running from 60-400MHz when computing the SECOND network on KITTI dataset. This work has been included by ISSCC2023 [1]. A demonstration is given to show the real-time 3D processing with a lidar sensor.
- 11 June 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-2
A Demonstration Platform for Large-Scaled Point Cloud Network Based on 28nm 2D/3D Unified Sparse Convolution Accelerator
3D point cloud processing plays an important role in many emerging applications such as autonomous driving, visual navigation, and virtual reality. It calls for hardware acceleration of multiple key operations, including 3D Submanifold SCONV, 3D non-Submanifold SCONV, and 2D SCONV. This work presents a 2D/3D unified sparse convolution accelerator for large-scale voxel-based point cloud networks. The chip is fabricated in TSMC 28nm CMOS technology to achieve 3.3-16.9 FPS running from 60-400MHz when computing the SECOND network on KITTI dataset. This work has been included by ISSCC2023 [1]. A demonstration is given to show the real-time 3D processing with a lidar sensor.
- 1 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 57 (8) , 2574-2585
A 65-nm Energy-Efficient Interframe Data Reuse Neural Network Accelerator for Video Applications
An energy-efficient convolutional neural network (CNN) accelerator is proposed for the video application. Previous works exploited the sparsity of differential (Diff) frame activation, but the improvement is limited as many Diff-frame data is small but non-zero. Processing of irregular sparse data also leads to low hardware utilization. To solve these problems, two key innovations are proposed in this article. First, we implement a hybrid-precision inter-frame-reuse architecture which takes advantage of both low bit-width and high sparsity of Diff-frame data. This technology can accelerate 3.2$\times $ inference speed with no accuracy loss. Second, we design a conv-pattern-aware processing array that achieves the 2.48$\times $ –14.2$\times $ PE utilization rate to process sparse data for different convolution kernels. The accelerator chip was implemented in 65-nm CMOS technology. To the best of our knowledge, it is the first silicon-proven CNN accelerator that supports inter-frame data reuse. Attributed to the inter-frame similarity, this video CNN accelerator reaches the minimum energy consumption of 24.7$\mu \text{J}$ /frame in the MobileNet-slim model, which is 76.3% less than the baseline.- 1 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 57 (8) , 2574-2585
A 65-nm Energy-Efficient Interframe Data Reuse Neural Network Accelerator for Video Applications
An energy-efficient convolutional neural network (CNN) accelerator is proposed for the video application. Previous works exploited the sparsity of differential (Diff) frame activation, but the improvement is limited as many Diff-frame data is small but non-zero. Processing of irregular sparse data also leads to low hardware utilization. To solve these problems, two key innovations are proposed in this article. First, we implement a hybrid-precision inter-frame-reuse architecture which takes advantage of both low bit-width and high sparsity of Diff-frame data. This technology can accelerate 3.2$\times $ inference speed with no accuracy loss. Second, we design a conv-pattern-aware processing array that achieves the 2.48$\times $ –14.2$\times $ PE utilization rate to process sparse data for different convolution kernels. The accelerator chip was implemented in 65-nm CMOS technology. To the best of our knowledge, it is the first silicon-proven CNN accelerator that supports inter-frame data reuse. Attributed to the inter-frame similarity, this video CNN accelerator reaches the minimum energy consumption of 24.7$\mu \text{J}$ /frame in the MobileNet-slim model, which is 76.3% less than the baseline.- 4 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 55 (2) , 465-477
STICKER: An Energy-Efficient Multi-Sparsity Compatible Accelerator for Convolutional Neural Networks in 65-nm CMOS
STICKER is an energy-efficient convolutional neural network (NN) processor. It mainly improves energy efficiency by making full use of sparsity. The network sparsity can potentially lower storage and computation requirements. However, the sparsity distribution of both activations and weights ranges from 2% to 99% in different layers or models. Therefore, good support for the sparsity distribution is the key to improve the energy efficiency. Three new features are proposed in this article to support wide sparsity distribution efficiently. First, multi-sparsity control and data flow are implemented for finer sparsity granularity support. It can automatically switch the processor among nine sparsity modes for higher energy efficiency. Second, a multi-mode hierarchical data memory which can be reconfigured for networks with different sparsity modes is designed for higher storage efficiency. Third, a multi-sparsity-compatible set-associative convolution processing element (PE) array is designed to efficiently carry out convolution operations under different sparsity modes, especially when both activations and weights are sparse. STICKER was implemented in a 65-nm CMOS technology. With its wide-range sparsity-supported capacity, the peak energy efficiency reaches 62.1 TOPS/W when sparsity ratios of both activations and weights are 5%. In a completely pruned Alexnet model, STICKER achieves 2.82 TOPS/W energy efficiency 1.8x higher than that of the state-of-the-art processors.
- 4 November 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 55 (2) , 465-477
STICKER: An Energy-Efficient Multi-Sparsity Compatible Accelerator for Convolutional Neural Networks in 65-nm CMOS
STICKER is an energy-efficient convolutional neural network (NN) processor. It mainly improves energy efficiency by making full use of sparsity. The network sparsity can potentially lower storage and computation requirements. However, the sparsity distribution of both activations and weights ranges from 2% to 99% in different layers or models. Therefore, good support for the sparsity distribution is the key to improve the energy efficiency. Three new features are proposed in this article to support wide sparsity distribution efficiently. First, multi-sparsity control and data flow are implemented for finer sparsity granularity support. It can automatically switch the processor among nine sparsity modes for higher energy efficiency. Second, a multi-mode hierarchical data memory which can be reconfigured for networks with different sparsity modes is designed for higher storage efficiency. Third, a multi-sparsity-compatible set-associative convolution processing element (PE) array is designed to efficiently carry out convolution operations under different sparsity modes, especially when both activations and weights are sparse. STICKER was implemented in a 65-nm CMOS technology. With its wide-range sparsity-supported capacity, the peak energy efficiency reaches 62.1 TOPS/W when sparsity ratios of both activations and weights are 5%. In a completely pruned Alexnet model, STICKER achieves 2.82 TOPS/W energy efficiency 1.8x higher than that of the state-of-the-art processors.
- 7 January 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 54 (3) , 885-895
A 130-nm Ferroelectric Nonvolatile System-on-Chip With Direct Peripheral Restore Architecture for Transient Computing System
Owing to its unique capability to sustain computation progress over power outages, a nonvolatile processor (NVP) is promising for energy-harvesting-powered Internet-of-Things devices. However, the widespread application of NVP is continually blocked by the system integration issues and the configuration overheads of peripheral devices. This paper presents a nonvolatile system-on-chip (NVSoC) with improved integration level, power management flexibility, and system wake-up speed. An on-chip power management subsystem is designed to minimize the number of external components while supporting versatile power policies. And a direct peripheral restore architecture is outlined, which enables a fast and parallel re-configuration of peripheral devices after the resumption of power supply. A test chip is fabricated in a 130-nm ferroelectric-CMOS process with 22.09-mm² area. Measurement results show 6x higher data throughput as compared with a conventional NVP when facing power failures.
- 7 January 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. 54 (3) , 885-895
A 130-nm Ferroelectric Nonvolatile System-on-Chip With Direct Peripheral Restore Architecture for Transient Computing System
Owing to its unique capability to sustain computation progress over power outages, a nonvolatile processor (NVP) is promising for energy-harvesting-powered Internet-of-Things devices. However, the widespread application of NVP is continually blocked by the system integration issues and the configuration overheads of peripheral devices. This paper presents a nonvolatile system-on-chip (NVSoC) with improved integration level, power management flexibility, and system wake-up speed. An on-chip power management subsystem is designed to minimize the number of external components while supporting versatile power policies. And a direct peripheral restore architecture is outlined, which enables a fast and parallel re-configuration of peripheral devices after the resumption of power supply. A test chip is fabricated in a 130-nm ferroelectric-CMOS process with 22.09-mm² area. Measurement results show 6x higher data throughput as compared with a conventional NVP when facing power failures.
- 26 October 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 66 (5) , 1970-1980
A 2.2-GHz Configurable Direct Digital Frequency Synthesizer Based on LUT and Rotation
This paper proposes a configurable direct digital frequency synthesizer (DDFS) based on the lookup-table (LUT)-rotation architecture. To break through the limitation of the single mode, the proposed DDFS is the first attempt which supports four-mode switching on chip. Taking the advantages of small LUT size and pipelined rotation, the DDFS achieves both high-speed and high-resolution properties. The multi-bit rotation tree is helpful to reduce the latency and register usage, which improves the agility and energy efficiency of the DDFS. A partition problem is raised in this paper, and we estimate the optimal solution through both the theory and experiment. Based on the partition, an output can achieve the highest spurious-free dynamic range (SFDR) with as small as possible LUT size. The functionality of the proposed DDFS has been validated in a Xilinx ZedBoard field-programmable gate array. The synthesis and layout results show that the chip achieves maximum 102-dBc SFDR and 2.2-GHz clock frequency. The power consumption and latency cycles reach minimum 6.9 mW/GHz and 7 cycles, which are 20% and 30% reduction compared with the state-of-the-art DDFS.
- 26 October 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 66 (5) , 1970-1980
A 2.2-GHz Configurable Direct Digital Frequency Synthesizer Based on LUT and Rotation
This paper proposes a configurable direct digital frequency synthesizer (DDFS) based on the lookup-table (LUT)-rotation architecture. To break through the limitation of the single mode, the proposed DDFS is the first attempt which supports four-mode switching on chip. Taking the advantages of small LUT size and pipelined rotation, the DDFS achieves both high-speed and high-resolution properties. The multi-bit rotation tree is helpful to reduce the latency and register usage, which improves the agility and energy efficiency of the DDFS. A partition problem is raised in this paper, and we estimate the optimal solution through both the theory and experiment. Based on the partition, an output can achieve the highest spurious-free dynamic range (SFDR) with as small as possible LUT size. The functionality of the proposed DDFS has been validated in a Xilinx ZedBoard field-programmable gate array. The synthesis and layout results show that the chip achieves maximum 102-dBc SFDR and 2.2-GHz clock frequency. The power consumption and latency cycles reach minimum 6.9 mW/GHz and 7 cycles, which are 20% and 30% reduction compared with the state-of-the-art DDFS.
- 1 January 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
A 2-GHz Direct Digital Frequency Synthesizer Based on LUT and Rotation
This paper proposes a direct digital frequency synthesizer (DDFS) based on Lookup-Table-Rotation (LUT-ROT) architecture. The DDFS takes the advantages of coarse-fine LUT and pipelined rotation units to achieve both high-speed and high-resolution. Based on the analysis of approximation error, the trade-off between accuracy and memory usage is achieved, which leads to the lowest amplitude of noise and the smallest size of LUT. Experimental results show that the DDFS achieves 11.7mW/GHz power consumption and 96dBc SFDR in 2-GHz clock frequency.
- 1 January 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-5
A 2-GHz Direct Digital Frequency Synthesizer Based on LUT and Rotation
This paper proposes a direct digital frequency synthesizer (DDFS) based on Lookup-Table-Rotation (LUT-ROT) architecture. The DDFS takes the advantages of coarse-fine LUT and pipelined rotation units to achieve both high-speed and high-resolution. Based on the analysis of approximation error, the trade-off between accuracy and memory usage is achieved, which leads to the lowest amplitude of noise and the smallest size of LUT. Experimental results show that the DDFS achieves 11.7mW/GHz power consumption and 96dBc SFDR in 2-GHz clock frequency.