Deep Convolutional Neural Network Architecture With Reconfigurable Computation Patterns

Tu, Fengbin; Yin, Shouyi; Ouyang, Peng; Tang, Shibin; Liu, Leibo; Wei, Shaojun

doi:10.1109/tvlsi.2017.2688340

Cited by 264 publications

(133 citation statements)

References 11 publications

Supporting

Mentioning

130

Contrasting

Unclassified

Order By: Relevance

“…The DCNN classifier 38 gains significance over the other classifiers because of its ability to process the input image to perform the effective classification. The difference of DCNN with respect to the neural network (NN) is with regard to the arrangement of the neurons.…”

Section: Structure Of the Dcnnmentioning

confidence: 99%

Multiclassifier for severity‐level categorization of glioma tumors using multimodal magnetic resonance imaging brain images

Mahesh

Renjit²

2019

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Brain tumor segmentation and classification is a crucial challenge in diagnosing, planning, and treating brain tumors. This article proposes an automatic method that categorizes the severity level of the tumors to render an effective diagnosis. The proposed fractional Jaya optimizer‐deep convolutional neural network undergoes the severity classification based on the features obtained from the segments of the magnetic resonance imaging (MRI) images. The segments are obtained using the particle swarm optimization that ensures the optimal selection of the segments from the MRI image and yields the core tumor and the edema tumor regions. The experimentation using the BRATS database reveals that the proposed method acquired a maximal accuracy, specificity, and sensitivity of 0.9414, 0.9429, and 0.9708, respectively.

show abstract

Section: Structure Of the Dcnnmentioning

confidence: 99%

Multiclassifier for severity‐level categorization of glioma tumors using multimodal magnetic resonance imaging brain images

Mahesh

Renjit²

2019

Int J Imaging Syst Tech

View full text Add to dashboard Cite

show abstract

“…根据这些特性, 我们提出了通用性的解决方案 --一款高效的可重构神经计算芯片 [10] (代号 "Thinker" 传统的卷积神经网络加速器如 Eyeriss [3] , Envision [11] 针对卷积做了很多研究如数据复用或并行计算, 但是针对反卷积模块却有着极低的 PE 利用率. 我们基于 Thinker 的架构模型, 提出了针对生成网络模型的可重构硬件架构 [12] (代号 "GNA" 神经网络加速器执行过程中, 对片上缓存和片外 RAM 的访问问题, 并提出相应的数据复用技术 [13] .…”

Section: 针对混合神经网络的可重构计算芯片架构unclassified

Research on low-power neural network computing accelerator

Yan

Zhang

et al. 2019

Sci. Sin.-Inf.

Self Cite

View full text Add to dashboard Cite

A novel voltage-type sense amplifier for low-power nonvolatile memories SCIENCE CHINA Information Sciences 53, 1676 (2010); Low-power design and application based on CSD optimization for a fixed coefficient multiplier SCIENCE CHINA Information Sciences 54, 2443 (2011); A low-jitter low-power monolithically integrated optical receiver for SDH STM-16 SCIENCE CHINA Information Sciences 54, 1293 (2011); A novel compact low-power direct conversion receiver for mobile UHF RFID reader

show abstract

“…We now propose changes to the Morph base architecture so that it can better match the needs of different 3D CNN layers. In general, there are four aspects to imparting flexibility to rigid CNN accelerators, namely configurable buffers, PE control logic, NoCs [38] and datapaths [35], [39]. We will add support for flexible buffer partitions to enable different tile sizes per 3D CNN data type without introducing buffer fragmentation.…”

Section: B Morph: a Flexible Architecturementioning

confidence: 99%

Morph: Flexible Acceleration for 3D CNN-Based Video Understanding

Hegde

Agrawal

Yao

et al. 2018

2018 51st Annual IEEE/ACM International Symposium on Microarchitecture (MICRO)

View full text Add to dashboard Cite

The past several years have seen both an explosion in the use of Convolutional Neural Networks (CNNs) and the design of accelerators to make CNN inference practical. In the architecture community, the lion share of effort has targeted CNN inference for image recognition. The closely related problem of video recognition has received far less attention as an accelerator target. This is surprising, as video recognition is more computationally intensive than image recognition, and video traffic is predicted to be the majority of internet traffic in the coming years.This paper fills the gap between algorithmic and hardware advances for video recognition by providing a design space exploration and flexible architecture for accelerating 3D Convolutional Neural Networks (3D CNNs)-the core kernel in modern video understanding. When compared to (2D) CNNs used for image recognition, efficiently accelerating 3D CNNs poses a significant engineering challenge due to their large (and variable over time) memory footprint and higher dimensionality.To address these challenges, we design a novel accelerator, called Morph, that can adaptively support different spatial and temporal tiling strategies depending on the needs of each layer of each target 3D CNN. We codesign a software infrastructure alongside the Morph hardware to find good-fit parameters to control the hardware. Evaluated on state-of-the-art 3D CNNs, Morph achieves up to 3.4× (2.5× average) reduction in energy consumption and improves performance/watt by up to 5.1× (4× average) compared to a baseline 3D CNN accelerator, with an area overhead of 5%. Morph further achieves a 15.9× average energy reduction on 3D CNNs when compared to Eyeriss.

show abstract

Deep Convolutional Neural Network Architecture With Reconfigurable Computation Patterns

Cited by 264 publications

References 11 publications

Multiclassifier for severity‐level categorization of glioma tumors using multimodal magnetic resonance imaging brain images

Multiclassifier for severity‐level categorization of glioma tumors using multimodal magnetic resonance imaging brain images

Research on low-power neural network computing accelerator

Morph: Flexible Acceleration for 3D CNN-Based Video Understanding

Contact Info

Product

Resources

About