AI Chat AI Image Generator AI Video Text to Speech

Efficient Accelerator for Dilated and Transposed Convolution with Decomposition

05/02/2022

∙

by Kuo-Wei Chang, et al.

∙

∙

Hardware acceleration for dilated and transposed convolution enables real time execution of related tasks like segmentation, but current designs are specific for these convolutional types or suffer from complex control for reconfigurable designs. This paper presents a design that decomposes input or weight for dilated and transposed convolutions respectively to skip redundant computations and thus executes efficiently on existing dense CNN hardware as well. The proposed architecture can cut down 87.8% of the cycle counts to achieve 8.2X speedup over a naive execution for the ENet case.

Kuo-Wei Chang
4 publications
Tian-Sheuan Chang
22 publications

page 1

page 2

page 3

research

∙ 05/02/2022

VSCNN: Convolution Neural Network Accelerator With Vector Sparsity

Hardware accelerator for convolution neural network (CNNs) enables real ...

0 Kuo-Wei Chang, et al. ∙

research

∙ 05/02/2022

VWA: Hardware Efficient Vectorwise Accelerator for Convolutional Neural Network

Hardware accelerators for convolution neural networks (CNNs) enable real...

0 Kuo-Wei Chang, et al. ∙

research

∙ 04/29/2021

Hardware Architecture of Embedded Inference Accelerator and Analysis of Algorithms for Depthwise and Large-Kernel Convolutions

In order to handle modern convolutional neural networks (CNNs) efficient...

0 Tse-Wei Chen, et al. ∙

research

∙ 05/02/2022

BSRA: Block-based Super Resolution Accelerator with Hardware Efficient Pixel Attention

Increasingly, convolution neural network (CNN) based super resolution mo...

0 Dun-Hao Yang, et al. ∙

research

∙ 02/26/2021

A Reconfigurable Winograd CNN Accelerator with Nesting Decomposition Algorithm for Computing Convolution with Large Filters

Recent literature found that convolutional neural networks (CNN) with la...

0 Jingbo Jiang, et al. ∙

research

∙ 05/31/2019

Efficient Multiway Hash Join on Reconfigurable Hardware

We propose the algorithms for performing multiway joins using a new type...

0 Kunle Olukotun, et al. ∙

research

∙ 10/28/2021

SIMCNN – Exploiting Computational Similarity to Accelerate CNN Training in Hardware

Convolution neural networks (CNN) are computation intensive to train. It...

0 Vahid Janfaza, et al. ∙