As of April 2, 2024, I’m reaching out to share my experience and seek advice or support regarding running YOLOv5 on the NVIDIA Jetson Nano. Despite extensive efforts over the past three days, including consulting GPT Enterprise, Stack Overflow, YouTube tutorials, and other resources, I have been unable to successfully deploy YOLOv5 on this platform using Python 3.6.9 and OpenCV with CUDA support.
My journey led me to discover that previous instructions and methods for installing YOLOv5 are now outdated and no longer applicable. This realization came after investing $157 in the Jetson Nano, motivated by its potential for developing a portable and cost-effective EDGE AI application.
Given the recent EOL announcement for the Jetson Nano Dev Kit, I fear that this challenge may not garner the attention it needs from the YOLOv5 developers, possibly due to the board’s aging status. However, the form factor and capabilities of the Jetson Nano made it an ideal choice for my project’s requirements.
I have already opened two tickets with the YOLOv5 repository, hoping for a resolution:
here is one of the tickets
here is the other:
At this juncture, I’m contemplating whether the Jetson Nano was the right choice for my project. I would deeply appreciate any guidance, support, or possible trade-in options that NVIDIA might offer towards acquiring a more suitable board for my needs, such as the Orion or another small form factor, yet more current device. My project aims to leverage EDGE AI in a portable and economically viable manner, and finding the right hardware platform is crucial.
To those in the community considering YOLOv5 or YOLOv9 for their Jetson Nano projects, I hope my experience can save you time and resources. Any advice, insights, or solutions from NVIDIA or the community would be invaluable and greatly appreciated.
Unless you see an update here, there is no solution or fix for YOLOv5 or YOLOv9 on Jetson Nano.
If you want YOLOv5, please try the Deesptream since it has been tested and optimized on Nano.
But if you want to use a newer model like YOLOv8 or YOLOv9.
You might need an environment that can support newer Python versions, like Xavier or Orin series.