We are testing with Jetson Nano Developer Kit(photo3) for Image Processing through Yolo V5.
Installed Jetpack 4.6.3 (however, when we checked with Jtop, it shows 4.6.4. Not a big deal it seems).
- Updated and upgraded.
Since the Ultralytics (for yolo v5), will not support python <3.8, we created a Virtual environment
Installed Python 3.8.
Since the default Opencv is 4.1.1 (with Cuda showing NO in Jtop) for Jetpack 4.6.3, we Installed OpenCV4.5.1
CMAKE file. Showed Python 3.8 and Lib files (NO). wrappers not found
then installed numpy and added the Lib files path in cMAKE script and did again.
Now the Python 3.8 showed Lib files path (Photo4)
Checked Cuda and it showed YES.
Then started our python program. The ultralytics version (8.0.180) we installed was generating some error.
Hence, we installed a lower version (8.0.113)
Now, the program started to run
However, the main issue is the GPU is not functional and the entire program works in CPU only.
Hence, each frame process takes around 2 seconds.
Even though torch vision was available in pip list check, we tried to install the the pack from the nvidia official website.
Received error (photo1 & photo2)
Our requirement.:
To enable GPU for image processing with Ultralytics Yolo trained custom model.
Request.
Can you please guide us where we are doing mistake.