Article ID: 000089046 Content Type: Compatibility Last Reviewed: 05/20/2022

Why Inferencing Using YOLOv3 on Ubuntu* 20.04 Is Faster than Inferencing on Yocto* OS?

Environment

Ubuntu 20.04 Yocto (IOTG Yocto BSP, Tag: release-74_tgl_u_mr3) Yocto (IOTG Yocto BSP, Tag: release-81_tgl_u_mr4) OpenVINO 2021.4 OpenVINO 2021.4.2

BUILT IN - ARTICLE INTRO SECOND COMPONENT
Summary

Correct version of TGL Yocto source for inferencing with OpenVINO™ 2021.4.2

Description
  • Ran Object Detection C++ Demo with YOLOv3 model using GPU plugin in OpenVINO™ 2021.4.
  • Obtained similar performances in Ubuntu 20.04 and Yocto (IOTG Yocto BSP, Tag: release-74_tgl_u_mr3) when using Intel® Core™ i5-1145G7E, Intel® Core™ i3-1115G4E, and Intel® Celeron® 6305E processors.
  • Obtained 58 frames per second (FPS) in Ubuntu 20.04 and 47.3 FPS in Yocto (IOTG Yocto BSP, Tag: release-74_tgl_u_mr3) when using Intel® Core™ i7-1185G7E processor.
Resolution
  • Use IOTG Yocto BSP (Tag: release-81_tgl_u_mr4) image for Yocto. 
  • Run Object Detection C++ Demo with YOLOv3 model using GPU plugin in OpenVINO™ 2021.4.2.

Related Products

This article applies to 2 products