Mengshu Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Mengshu Sun has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Xue Lin
Northeastern University
9 shared publications - Yanzhi Wang
Northeastern University
8 shared publications - Peiyan Dong
Department of Electrical and Computer Engineering, Northeastern University, Boston, MA, USA
5 shared publications - Wei Niu
University of Georgia, Athens, GA, USA
4 shared publications - Zhengang Li
Northeastern University, USA
4 shared publications - Zhenglun Kong
Department of Electrical and Computer Engineering, Northeastern University, Boston, USA
4 shared publications
Latest Publications
- Open Access
- 19 September 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22  (3) , 1-22
Mobile-3DCNN: An Acceleration Framework for Ultra-Real-Time Execution of Large 3D CNNs on Mobile Devices
It is challenging to deploy 3D Convolutional Neural Networks (3D CNNs) on mobile devices, specifically if both real-time execution and high inference accuracy are in demand, because the increasingly large model size and complex model structure of 3D CNNs usually require tremendous computation and memory resources. Weight pruning is proposed to mitigate this challenge. However, existing pruning is either not compatible with modern parallel architectures, resulting in long inference latency or subject to significant accuracy degradation. This article proposes an end-to-end 3D CNN acceleration framework based on pruning/compilation co-design called Mobile-3DCNN that consists of two parts: a novel, fine-grained structured pruning enhanced by a prune/Winograd adaptive selection (that is mobile-hardware-friendly and can achieve high pruning accuracy), and a set of compiler optimization and code generation techniques enabled by our pruning (to fully transform the pruning benefit to real performance gains). The evaluation demonstrates that Mobile-3DCNN outperforms state-of-the-art end-to-end DNN acceleration frameworks that support 3D CNN execution on mobile devices, Alibaba Mobile Neural Networks and Pytorch-Mobile with speedup up to 34× with minor accuracy degradation, proving it is possible to execute high-accuracy large 3D CNNs on mobile devices in real-time (or even ultra-real-time).
- 19 September 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Architecture and Code Optimization
- Vol. 22  (3) , 1-22
Open AccessMobile-3DCNN: An Acceleration Framework for Ultra-Real-Time Execution of Large 3D CNNs on Mobile Devices
It is challenging to deploy 3D Convolutional Neural Networks (3D CNNs) on mobile devices, specifically if both real-time execution and high inference accuracy are in demand, because the increasingly large model size and complex model structure of 3D CNNs usually require tremendous computation and memory resources. Weight pruning is proposed to mitigate this challenge. However, existing pruning is either not compatible with modern parallel architectures, resulting in long inference latency or subject to significant accuracy degradation. This article proposes an end-to-end 3D CNN acceleration framework based on pruning/compilation co-design called Mobile-3DCNN that consists of two parts: a novel, fine-grained structured pruning enhanced by a prune/Winograd adaptive selection (that is mobile-hardware-friendly and can achieve high pruning accuracy), and a set of compiler optimization and code generation techniques enabled by our pruning (to fully transform the pruning benefit to real performance gains). The evaluation demonstrates that Mobile-3DCNN outperforms state-of-the-art end-to-end DNN acceleration frameworks that support 3D CNN execution on mobile devices, Alibaba Mobile Neural Networks and Pytorch-Mobile with speedup up to 34× with minor accuracy degradation, proving it is possible to execute high-accuracy large 3D CNNs on mobile devices in real-time (or even ultra-real-time).
- Open Access
- 30 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 324-337
Quasar-ViT: Hardware-Oriented Quantization-Aware Architecture Search for Vision Transformers
Vision transformers (ViTs) have demonstrated their superior accuracy for computer vision tasks compared to convolutional neural networks (CNNs). However, ViT models are often computation-intensive for efficient deployment on resource-limited edge devices. This work proposes Quasar-ViT, a hardware-oriented quantization-aware architecture search framework for ViTs, to design efficient ViT models for hardware implementation while preserving the accuracy. First, Quasar-ViT trains a supernet using our row-wise flexible mixed-precision quantization scheme, mixed-precision weight entanglement, and supernet layer scaling techniques. Then, it applies an efficient hardware-oriented search algorithm, integrated with hardware latency and resource modeling, to determine a series of optimal subnets from supernet under different inference latency targets. Finally, we propose a series of model-adaptive designs on the FPGA platform to support the architecture search and mitigate the gap between the theoretical computation reduction and the practical inference speedup. Our searched models achieve 101.5, 159.6, and 251.6 frames-per-second (FPS) inference speed on the AMD/Xilinx ZCU102 FPGA with 80.4%, 78.6%, and 74.9% top-1 accuracy, respectively, for the ImageNet dataset, consistently outperforming prior works.
- 30 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 324-337
Open AccessQuasar-ViT: Hardware-Oriented Quantization-Aware Architecture Search for Vision Transformers
Vision transformers (ViTs) have demonstrated their superior accuracy for computer vision tasks compared to convolutional neural networks (CNNs). However, ViT models are often computation-intensive for efficient deployment on resource-limited edge devices. This work proposes Quasar-ViT, a hardware-oriented quantization-aware architecture search framework for ViTs, to design efficient ViT models for hardware implementation while preserving the accuracy. First, Quasar-ViT trains a supernet using our row-wise flexible mixed-precision quantization scheme, mixed-precision weight entanglement, and supernet layer scaling techniques. Then, it applies an efficient hardware-oriented search algorithm, integrated with hardware latency and resource modeling, to determine a series of optimal subnets from supernet under different inference latency targets. Finally, we propose a series of model-adaptive designs on the FPGA platform to support the architecture search and mitigate the gap between the theoretical computation reduction and the practical inference speedup. Our searched models achieve 101.5, 159.6, and 251.6 frames-per-second (FPS) inference speed on the AMD/Xilinx ZCU102 FPGA with 80.4%, 78.6%, and 74.9% top-1 accuracy, respectively, for the ImageNet dataset, consistently outperforming prior works.
- 16 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 43  (10) , 3027-3040
Hardware-Friendly 3-D CNN Acceleration With Balanced Kernel Group Sparsity
Being capable of extracting more information than 2-D convolutional neural networks (CNNs), 3-D CNNs have been playing a vital role in video analysis tasks like human action recognition, but their massive operations hinder the real-time execution on edge devices with constrained computation and memory resources. Although various model compression techniques have been applied to accelerate 2-D CNNs, there are rare efforts in investigating hardware-friendly pruning of 3-D CNNs and acceleration on customizable edge platforms like FPGAs. This work starts from proposing a kernel group row-column (KGRC) weight sparsity pattern, which is fine-grained to achieve high pruning ratios with negligible accuracy loss, and balanced across kernel groups to achieve high computation parallelism on hardware. The reweighted pruning algorithm for this sparsity is then presented and performed on 3-D CNNs, followed by quantization under different precisions. Along with model compression, FPGA-based accelerators with four modes are designed in support of the kernel group sparsity in multiple dimensions. The co-design framework of the pruning algorithm and the accelerator is tested on two representative 3-D CNNs, namely C3D and R(2+1)D, with the Xilinx ZCU102 FPGA platform for action recognition. The experimental results indicate that the accelerator implementation with the KGRC sparsity and 8-bit quantization achieves a good balance between the speedup and model accuracy, leading to acceleration ratios of$4.12\times $ for C3D and$3.85\times $ for R(2+1)D compared with the 16-bit baseline designs supporting only dense models.- 16 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
- Vol. 43  (10) , 3027-3040
Hardware-Friendly 3-D CNN Acceleration With Balanced Kernel Group Sparsity
Being capable of extracting more information than 2-D convolutional neural networks (CNNs), 3-D CNNs have been playing a vital role in video analysis tasks like human action recognition, but their massive operations hinder the real-time execution on edge devices with constrained computation and memory resources. Although various model compression techniques have been applied to accelerate 2-D CNNs, there are rare efforts in investigating hardware-friendly pruning of 3-D CNNs and acceleration on customizable edge platforms like FPGAs. This work starts from proposing a kernel group row-column (KGRC) weight sparsity pattern, which is fine-grained to achieve high pruning ratios with negligible accuracy loss, and balanced across kernel groups to achieve high computation parallelism on hardware. The reweighted pruning algorithm for this sparsity is then presented and performed on 3-D CNNs, followed by quantization under different precisions. Along with model compression, FPGA-based accelerators with four modes are designed in support of the kernel group sparsity in multiple dimensions. The co-design framework of the pruning algorithm and the accelerator is tested on two representative 3-D CNNs, namely C3D and R(2+1)D, with the Xilinx ZCU102 FPGA platform for action recognition. The experimental results indicate that the accelerator implementation with the KGRC sparsity and 8-bit quantization achieves a good balance between the speedup and model accuracy, leading to acceleration ratios of$4.12\times $ for C3D and$3.85\times $ for R(2+1)D compared with the 16-bit baseline designs supporting only dense models.- Open Access
- 26 June 2023
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 37  (7) , 8360-8368
Peeling the Onion: Hierarchical Reduction of Data Redundancy for Efficient Vision Transformer Training
Vision transformers (ViTs) have recently obtained success in many applications, but their intensive computation and heavy memory usage at both training and inference time limit their generalization. Previous compression algorithms usually start from the pre-trained dense models and only focus on efficient inference, while time-consuming training is still unavoidable. In contrast, this paper points out that the million-scale training data is redundant, which is the fundamental reason for the tedious training. To address the issue, this paper aims to introduce sparsity into data and proposes an end-to-end efficient training framework from three sparse perspectives, dubbed Tri-Level E-ViT. Specifically, we leverage a hierarchical data redundancy reduction scheme, by exploring the sparsity under three levels: number of training examples in the dataset, number of patches (tokens) in each example, and number of connections between tokens that lie in attention weights. With extensive experiments, we demonstrate that our proposed technique can noticeably accelerate training for various ViT architectures while maintaining accuracy. Remarkably, under certain ratios, we are able to improve the ViT accuracy rather than compromising it. For example, we can achieve 15.2% speedup with 72.6% (+0.4) Top-1 accuracy on Deit-T, and 15.7% speedup with 79.9% (+0.1) Top-1 accuracy on Deit-S. This proves the existence of data redundancy in ViT. Our code is released at https://github.com/ZLKong/Tri-Level-ViT
- 26 June 2023
- Published by Association for the Advancement of Artificial Intelligence (AAAI) in Proceedings of the AAAI Conference on Artificial Intelligence
- Vol. 37  (7) , 8360-8368
Open AccessPeeling the Onion: Hierarchical Reduction of Data Redundancy for Efficient Vision Transformer Training
Vision transformers (ViTs) have recently obtained success in many applications, but their intensive computation and heavy memory usage at both training and inference time limit their generalization. Previous compression algorithms usually start from the pre-trained dense models and only focus on efficient inference, while time-consuming training is still unavoidable. In contrast, this paper points out that the million-scale training data is redundant, which is the fundamental reason for the tedious training. To address the issue, this paper aims to introduce sparsity into data and proposes an end-to-end efficient training framework from three sparse perspectives, dubbed Tri-Level E-ViT. Specifically, we leverage a hierarchical data redundancy reduction scheme, by exploring the sparsity under three levels: number of training examples in the dataset, number of patches (tokens) in each example, and number of connections between tokens that lie in attention weights. With extensive experiments, we demonstrate that our proposed technique can noticeably accelerate training for various ViT architectures while maintaining accuracy. Remarkably, under certain ratios, we are able to improve the ViT accuracy rather than compromising it. For example, we can achieve 15.2% speedup with 72.6% (+0.4) Top-1 accuracy on Deit-T, and 15.7% speedup with 79.9% (+0.1) Top-1 accuracy on Deit-S. This proves the existence of data redundancy in ViT. Our code is released at https://github.com/ZLKong/Tri-Level-ViT
- Open Access
HeatViT: Hardware-Efficient Adaptive Token Pruning for Vision Transformers
While vision transformers (ViTs) have continuously achieved new milestones in the field of computer vision, their sophisticated network architectures with high computation and memory costs have impeded their deployment on resource-limited edge devices. In this paper, we propose a hardware-efficient image-adaptive token pruning framework called HeatViT for efficient yet accurate ViT acceleration on embedded FPGAs. By analyzing the inherent computational patterns in ViTs, we first design an effective attention-based multi-head token selector, which can be progressively inserted before transformer blocks to dynamically identify and consolidate the non-informative tokens from input images. Moreover, we implement the token selector on hardware by adding miniature control logic to heavily reuse existing hardware components built for the backbone ViT. To improve the hardware efficiency, we further employ 8-bit fixed-point quantization, and propose polynomial approximations with regularization effect on quantization error for the frequently used nonlinear functions in ViTs. Finally, we propose a latency-aware multi-stage training strategy to determine the transformer blocks for inserting token selectors and optimize the desired (average) pruning rates for inserted token selectors, in order to improve both the model accuracy and inference latency on hardware. Compared to existing ViT pruning studies, under the similar computation cost, HeatViT can achieve 0.7%$\sim$8.9% higher accuracy; while under the similar model accuracy, HeatViT can achieve more than 28.4%$\sim$65.3% computation reduction, for various widely used ViTs, including DeiT-T, DeiT-S, DeiT-B, LV-ViT-S, and LV-ViT-M, on the ImageNet dataset. Compared to the baseline hardware accelerator, our implementations of HeatViT on the Xilinx ZCU102 FPGA achieve 3.46$\times$$\sim$4.89$\times$ speedup.
Open AccessHeatViT: Hardware-Efficient Adaptive Token Pruning for Vision Transformers
While vision transformers (ViTs) have continuously achieved new milestones in the field of computer vision, their sophisticated network architectures with high computation and memory costs have impeded their deployment on resource-limited edge devices. In this paper, we propose a hardware-efficient image-adaptive token pruning framework called HeatViT for efficient yet accurate ViT acceleration on embedded FPGAs. By analyzing the inherent computational patterns in ViTs, we first design an effective attention-based multi-head token selector, which can be progressively inserted before transformer blocks to dynamically identify and consolidate the non-informative tokens from input images. Moreover, we implement the token selector on hardware by adding miniature control logic to heavily reuse existing hardware components built for the backbone ViT. To improve the hardware efficiency, we further employ 8-bit fixed-point quantization, and propose polynomial approximations with regularization effect on quantization error for the frequently used nonlinear functions in ViTs. Finally, we propose a latency-aware multi-stage training strategy to determine the transformer blocks for inserting token selectors and optimize the desired (average) pruning rates for inserted token selectors, in order to improve both the model accuracy and inference latency on hardware. Compared to existing ViT pruning studies, under the similar computation cost, HeatViT can achieve 0.7%$\sim$8.9% higher accuracy; while under the similar model accuracy, HeatViT can achieve more than 28.4%$\sim$65.3% computation reduction, for various widely used ViTs, including DeiT-T, DeiT-S, DeiT-B, LV-ViT-S, and LV-ViT-M, on the ImageNet dataset. Compared to the baseline hardware accelerator, our implementations of HeatViT on the Xilinx ZCU102 FPGA achieve 3.46$\times$$\sim$4.89$\times$ speedup.
- Top Cited Papers
- 3 November 2022
- Published by Springer Nature
- p. 620-640
SPViT: Enabling Faster Vision Transformers via Latency-Aware Soft Token Pruning
- 3 November 2022
- Published by Springer Nature
- p. 620-640
Top Cited PapersSPViT: Enabling Faster Vision Transformers via Latency-Aware Soft Token Pruning
- Open Access
- 30 September 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 21  (5) , 1-22
Mobile or FPGA? A Comprehensive Evaluation on Energy Efficiency and a Unified Optimization Framework
Efficient deployment of Deep Neural Networks (DNNs) on edge devices (i.e., FPGAs and mobile platforms) is very challenging, especially under a recent witness of the increasing DNN model size and complexity. Model compression strategies, including weight quantization and pruning, are widely recognized as effective approaches to significantly reduce computation and memory intensities, and have been implemented in many DNNs on edge devices. However, most state-of-the-art works focus on ad hoc optimizations, and there lacks a thorough study to comprehensively reveal the potentials and constraints of different edge devices when considering different compression strategies. In this article, we qualitatively and quantitatively compare the energy efficiency of FPGA-based and mobile-based DNN executions using mobile GPU and provide a detailed analysis. Based on the observations obtained from the analysis, we propose a unified optimization framework using block-based pruning to reduce the weight storage and accelerate the inference speed on mobile devices and FPGAs, achieving high hardware performance and energy-efficiency gain while maintaining accuracy.
- 30 September 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 21  (5) , 1-22
Open AccessMobile or FPGA? A Comprehensive Evaluation on Energy Efficiency and a Unified Optimization Framework
Efficient deployment of Deep Neural Networks (DNNs) on edge devices (i.e., FPGAs and mobile platforms) is very challenging, especially under a recent witness of the increasing DNN model size and complexity. Model compression strategies, including weight quantization and pruning, are widely recognized as effective approaches to significantly reduce computation and memory intensities, and have been implemented in many DNNs on edge devices. However, most state-of-the-art works focus on ad hoc optimizations, and there lacks a thorough study to comprehensively reveal the potentials and constraints of different edge devices when considering different compression strategies. In this article, we qualitatively and quantitatively compare the energy efficiency of FPGA-based and mobile-based DNN executions using mobile GPU and provide a detailed analysis. Based on the observations obtained from the analysis, we propose a unified optimization framework using block-based pruning to reduce the weight storage and accelerate the inference speed on mobile devices and FPGAs, achieving high hardware performance and energy-efficiency gain while maintaining accuracy.
- 6 October 2021
- Published by Springer Nature in Electrical Engineering
- Vol. 104  (3) , 1459-1471
Power management in hybrid electric vehicles using deep recurrent reinforcement learning
- 6 October 2021
- Published by Springer Nature in Electrical Engineering
- Vol. 104  (3) , 1459-1471
Power management in hybrid electric vehicles using deep recurrent reinforcement learning
- Open Access
RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices
Mobile devices are becoming an important carrier for deep learning tasks, as they are being equipped with powerful, high-end mobile CPUs and GPUs. However, it is still a challenging task to execute 3D Convolutional Neural Networks (CNNs) targeting for real-time performance, besides high inference accuracy. The reason is more complex model structure and higher model dimensionality overwhelm the available computation/storage resources on mobile devices. A natural way may be turning to deep learning weight pruning techniques. However, the direct generalization of existing 2D CNN weight pruning methods to 3D CNNs is not ideal for fully exploiting mobile parallelism while achieving high inference accuracy. This paper proposes RT3D, a model compression and mobile acceleration framework for 3D CNNs, seamlessly integrating neural network weight pruning and compiler code generation techniques. We propose and investigate two structured sparsity schemes i.e., the vanilla structured sparsity and kernel group structured (KGS) sparsity that are mobile acceleration friendly. The vanilla sparsity removes whole kernel groups, while KGS sparsity is a more fine-grained structured sparsity that enjoys higher flexibility while exploiting full on-device parallelism. We propose a reweighted regularization pruning algorithm to achieve the proposed sparsity schemes. The inference time speedup due to sparsity is approaching the pruning rate of the whole model FLOPs (floating point operations). RT3D demonstrates up to 29.1$\times$ speedup in end-to-end inference time comparing with current mobile frameworks supporting 3D CNNs, with moderate 1%-1.5% accuracy loss. The end-to-end inference time for 16 video frames could be within 150 ms, when executing representative C3D and R(2+1)D models on a cellphone. For the first time, real-time execution of 3D CNNs is achieved on off-the-shelf mobiles.
Open AccessRT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices
Mobile devices are becoming an important carrier for deep learning tasks, as they are being equipped with powerful, high-end mobile CPUs and GPUs. However, it is still a challenging task to execute 3D Convolutional Neural Networks (CNNs) targeting for real-time performance, besides high inference accuracy. The reason is more complex model structure and higher model dimensionality overwhelm the available computation/storage resources on mobile devices. A natural way may be turning to deep learning weight pruning techniques. However, the direct generalization of existing 2D CNN weight pruning methods to 3D CNNs is not ideal for fully exploiting mobile parallelism while achieving high inference accuracy. This paper proposes RT3D, a model compression and mobile acceleration framework for 3D CNNs, seamlessly integrating neural network weight pruning and compiler code generation techniques. We propose and investigate two structured sparsity schemes i.e., the vanilla structured sparsity and kernel group structured (KGS) sparsity that are mobile acceleration friendly. The vanilla sparsity removes whole kernel groups, while KGS sparsity is a more fine-grained structured sparsity that enjoys higher flexibility while exploiting full on-device parallelism. We propose a reweighted regularization pruning algorithm to achieve the proposed sparsity schemes. The inference time speedup due to sparsity is approaching the pruning rate of the whole model FLOPs (floating point operations). RT3D demonstrates up to 29.1$\times$ speedup in end-to-end inference time comparing with current mobile frameworks supporting 3D CNNs, with moderate 1%-1.5% accuracy loss. The end-to-end inference time for 16 video frames could be within 150 ms, when executing representative C3D and R(2+1)D models on a cellphone. For the first time, real-time execution of 3D CNNs is achieved on off-the-shelf mobiles.
- Open Access
- 13 May 2019
- Published by Association for Computing Machinery (ACM)
- p. 81-86
HSIM-DNN
Deep learning that utilizes large-scale deep neural networks (DNNs) is effective in automatic high-level feature extraction but also computation and memory intensive. Constructing DNNs using block-circulant matrices can simultaneously achieve hardware acceleration and model compression while maintaining high accuracy. This paper proposes HSIM-DNN, an accurate hardware simulator on the C++ platform, to simulate the exact behavior of DNN hardware implementations and thereby facilitate the block-circulant matrix-based design of DNN training and inference procedures in hardware. Real FPGA implementations validate the simulator with various circulant block sizes and data bit lengths taking into account accuracy, compression ratio and power consumption, which provides excellent insights for hardware design.
- 13 May 2019
- Published by Association for Computing Machinery (ACM)
- p. 81-86
Open AccessHSIM-DNN
Deep learning that utilizes large-scale deep neural networks (DNNs) is effective in automatic high-level feature extraction but also computation and memory intensive. Constructing DNNs using block-circulant matrices can simultaneously achieve hardware acceleration and model compression while maintaining high accuracy. This paper proposes HSIM-DNN, an accurate hardware simulator on the C++ platform, to simulate the exact behavior of DNN hardware implementations and thereby facilitate the block-circulant matrix-based design of DNN training and inference procedures in hardware. Real FPGA implementations validate the simulator with various circulant block sizes and data bit lengths taking into account accuracy, compression ratio and power consumption, which provides excellent insights for hardware design.