English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Computer vision performances have been significantly improved in recent yearsby Convolutional Neural Networks(CNN). Currently, applications using CNNalgorithms are deployed mainly on general purpose hardwares, such as CPUs, GPUsor FPGAs. However, power consumption, speed, accuracy, memory footprint, anddie size should all be taken into consideration for mobile and embeddedapplications. Domain Specific Architecture (DSA) for CNN is the efficient andpractical solution for CNN deployment and implementation. We designed andproduced a 28nm Two-Dimensional CNN-DSA accelerator with an ultrapower-efficient performance of 9.3TOPS/Watt and with all processing done in theinternal memory instead of outside DRAM. It classifies 224x224 RGB image inputsat more than 140fps with peak power consumption at less than 300mW and anaccuracy comparable to the VGG benchmark. The CNN-DSA accelerator isreconfigurable to support CNN model coefficients of various layer sizes andlayer types, including convolution, depth-wise convolution, short-cutconnections, max pooling, and ReLU. Furthermore, in order to better supportreal-world deployment for various application scenarios, especially withlow-end mobile and embedded platforms and MCUs (Microcontroller Units), we alsodesigned algorithms to fully utilize the CNN-DSA accelerator efficiently byreducing the dependency on external accelerator computation resources,including implementation of Fully-Connected (FC) layers within the acceleratorand compression of extracted features from the CNN-DSA accelerator. Live demoswith our CNN-DSA accelerator on mobile and embedded systems show itscapabilities to be widely and practically applied in the real world.

Ultra Power-Efficient CNN Domain Specific Accelerator with 9.3TOPS/Watt for Mobile and Embedded Applications