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

For Human Action Recognition tasks (HAR), 3D Convolutional Neural Networkshave proven to be highly effective, achieving state-of-the-art results. Thisstudy introduces a novel streaming architecture based toolflow for mapping suchmodels onto FPGAs considering the model's inherent characteristics and thefeatures of the targeted FPGA device. The HARFLOW3D toolflow takes as input a3D CNN in ONNX format and a description of the FPGA characteristics, generatinga design that minimizes the latency of the computation. The toolflow iscomprised of a number of parts, including i) a 3D CNN parser, ii) a performanceand resource model, iii) a scheduling algorithm for executing 3D models on thegenerated hardware, iv) a resource-aware optimization engine tailored for 3Dmodels, v) an automated mapping to synthesizable code for FPGAs. The ability ofthe toolflow to support a broad range of models and devices is shown through anumber of experiments on various 3D CNN and FPGA system pairs. Furthermore, thetoolflow has produced high-performing results for 3D CNN models that have notbeen mapped to FPGAs before, demonstrating the potential of FPGA-based systemsin this space. Overall, HARFLOW3D has demonstrated its ability to delivercompetitive latency compared to a range of state-of-the-art hand-tunedapproaches being able to achieve up to 5$\times$ better performance compared tosome of the existing works.

HARFLOW3D: A Latency-Oriented 3D-CNN Accelerator Toolflow for HAR on FPGA Devices