Skip to content

Experimental implementation of real-time object detection algorithm YOLOR on embedded systems (edge computing devices)

Notifications You must be signed in to change notification settings

ewth/yolor-edge

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

yolor-edge

An implementation of YOLOR running successfully on the NVIDIA Jetson Xavier NX edge computing device.

YOLOR

You Only Learn One Representation (YOLOR) is a novel, state-of-the-art object detection algorithm published in May 2021, and producing world-leading performance results.

All credit goes to:

This Code

YOLOR was published with an official implementation built on PyTorch (see below), built and tested on a PC. This code represents a replication of the implementation for running on an edge device, with some additions for flair.

Thesis

This code forms part of a B.Eng(Hons) (Instrumentation and Control / Industrial Computer Systems) thesis, with the aim of evaluating state-of-the-art artificial intelligence research for application in urban search and rescue. As successful implementation and testing of YOLOR on edge devices (embedded systems) was previously unavailable in literature, the overarching aim was assessing suitability for real-world field use in autonomous mobile robotics.

Releases at specific points are available below.

Further Details

The thesis this repository is attached to is slated for publication in February 2022. Further details, including more specifics around implementation and use, will be made available as soon as practical.

Commit History

Unfortunately, a fair bit of burning the midnight oil and running purely on caffeine led to some kafuffles around commit history in this repo, namely due to tired eyes mishmashing Git submodules.