ABM-SpConv-SIMD: Accelerating Convolutional Neural Network Inference for Industrial IoT Applications on Edge Devices

Li, Xianduo; Gong, Xiaoli; Wang, Dong; Zhang, Jin; Baker, Thar; Zhou, June; Lu, Tingjuan

doi:10.1109/tnse.2022.3154412

Cited by 10 publications

(6 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, it is difficult to find the same elements in a small kernel, while small kernels are the trend. The authors of [14,19] have confirmed this issue. The authors expanded the scope of sharing the same weights into several kernels to implement ABM-SpConv, which increases the complexity of the hardware architecture.…”

Section: Related Workmentioning

confidence: 71%

“…Another type of method focuses on reducing the use of multiplier units in the fil-ter_loop through the optimized design of the hardware, such as ABM-SpConv [9,14,19], MF-Conv [13], and the weight-sharing (WS) technique [20][21][22]. The key to ABM-SpConv [9] is to perform the multiplication operations and accumulation operations of convolution in two separate stages.…”

Section: Related Workmentioning

confidence: 99%

“…However, Table 1 implies that the WRA still relies more on DSP resources than on logic resources. ABM-SpConv [9,14,19] and MF-Conv [13] focus on reducing the use of multiplier units in the filter_loop, as shown in Figure 1. ABM-SpConv uses weights to group the input data, and the data corresponding to the same weight enter the same DSP.…”

Section: Motivationmentioning

confidence: 99%

See 2 more Smart Citations

WRA-MF: A Bit-Level Convolutional-Weight-Decomposition Approach to Improve Parallel Computing Efficiency for Winograd-Based CNN Acceleration

Xiang,

Lv,

Meng

et al. 2023

Electronics

View full text Add to dashboard Cite

FPGA-based convolutional neural network (CNN) accelerators have been extensively studied recently. To exploit the parallelism of multiplier–accumulator computation in convolution, most FPGA-based CNN accelerators heavily depend on the number of on-chip DSP blocks in the FPGA. Consequently, the performance of the accelerators is restricted by the limitation of the DSPs, leading to an imbalance in the utilization of other FPGA resources. This work proposes a multiplication-free convolutional acceleration scheme (named WRA-MF) to relax the pressure on the required DSP resources. Firstly, the proposed WRA-MF employs the Winograd algorithm to reduce the computational density, and it then performs bit-level convolutional weight decomposition to eliminate the multiplication operations. Furthermore, by extracting common factors, the complexity of the addition operations is reduced. Experimental results on the Xilinx XCVU9P platform show that the WRA-MF can achieve 7559 GOP/s throughput at a 509 MHz clock frequency for VGG16. Compared with state-of-the-art works, the WRA-MF achieves up to a 3.47×–27.55× area efficiency improvement. The results indicate that the proposed architecture achieves a high area efficiency while ameliorating the imbalance in the resource utilization.

show abstract

Section: Related Workmentioning

confidence: 71%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

WRA-MF: A Bit-Level Convolutional-Weight-Decomposition Approach to Improve Parallel Computing Efficiency for Winograd-Based CNN Acceleration

Xiang,

Lv,

Meng

et al. 2023

Electronics

View full text Add to dashboard Cite

show abstract

“…The utilization of Raspberry Pi extended beyond security applications. For example, ABM-SpConv-SIMD, an on-device inference optimization framework, was proposed by [25]. Implemented on Raspberry Pi devices, this framework aimed to accelerate network inference by fully utilizing the low-cost and common CPU resources.…”

Section: Raspberry Pimentioning

confidence: 99%

Efficient Server Room Monitoring with Flutterand Firebase - Real-Time Insights and InstantNotifications

Özkurt

2024

Preprint

View full text Add to dashboard Cite

A comprehensive server room monitoring system is presented in this research,leveraging Raspberry Pi devices, DHT11 temperature/humidity sensors, PIRmotion sensors, and magnetic reed switches. Integration of Flutter and Firebasetechnologies enabled real-time environmental monitoring, security event detection, and remote access control. The effectiveness of Raspberry Pi devices infacilitating cost-effective and scalable monitoring solutions was demonstrated,while Flutter and Firebase enhanced accessibility, usability, and responsiveness.The aim of this research is to improve server room management, security, andoperational efficiency, contributing to a more secure and resilient infrastructurein the digital age

show abstract

“…Researchers also showed an improved latency by optimizing graphical unit communication during computing for image classification tasks [50]. Scientists proposed a new framework using an acceleration approach to improve inference latency, known as ABM-SpConv-SIMD [51].…”

Section: Transformers In Computer Visionmentioning

confidence: 99%

Optimal Topology of Vision Transformer for Real-Time Video Action Recognition in an End-To-End Cloud Solution

Sarraf,

Kabia

2023

MAKE

View full text Add to dashboard Cite

This study introduces an optimal topology of vision transformers for real-time video action recognition in a cloud-based solution. Although model performance is a key criterion for real-time video analysis use cases, inference latency plays a more crucial role in adopting such technology in real-world scenarios. Our objective is to reduce the inference latency of the solution while admissibly maintaining the vision transformer’s performance. Thus, we employed the optimal cloud components as the foundation of our machine learning pipeline and optimized the topology of vision transformers. We utilized UCF101, including more than one million action recognition video clips. The modeling pipeline consists of a preprocessing module to extract frames from video clips, training two-dimensional (2D) vision transformer models, and deep learning baselines. The pipeline also includes a postprocessing step to aggregate the frame-level predictions to generate the video-level predictions at inference. The results demonstrate that our optimal vision transformer model with an input dimension of 56 × 56 × 3 with eight attention heads produces an F1 score of 91.497% for the testing set. The optimized vision transformer reduces the inference latency by 40.70%, measured through a batch-processing approach, with a 55.63% faster training time than the baseline. Lastly, we developed an enhanced skip-frame approach to improve the inference latency by finding an optimal ratio of frames for prediction at inference, where we could further reduce the inference latency by 57.15%. This study reveals that the vision transformer model is highly optimizable for inference latency while maintaining the model performance.

show abstract

ABM-SpConv-SIMD: Accelerating Convolutional Neural Network Inference for Industrial IoT Applications on Edge Devices

Cited by 10 publications

References 52 publications

WRA-MF: A Bit-Level Convolutional-Weight-Decomposition Approach to Improve Parallel Computing Efficiency for Winograd-Based CNN Acceleration

WRA-MF: A Bit-Level Convolutional-Weight-Decomposition Approach to Improve Parallel Computing Efficiency for Winograd-Based CNN Acceleration

Efficient Server Room Monitoring with Flutterand Firebase - Real-Time Insights and InstantNotifications

Optimal Topology of Vision Transformer for Real-Time Video Action Recognition in an End-To-End Cloud Solution

Contact Info

Product

Resources

About