How to do prediction and testing of the pre-trained segmentation models on personal system other than Jetson using opencv python script?

Hi @dusty_nv

I want to load and do prediction using pre-trained segmentation models on jetson inference repo such as FCN-ResNet18-MHP, FCN-ResNet18-DeepScene on personal system with python script e.g. test.py before using on Jetson.

I have referred the below reference for custom segmentation training implementation:

but not able to do inference on test image or .mp4 file

How to do prediction on image or .mp4 file using opencv python script on personal system?

Please let me know if any further information w.r.t is needed.

Hi @nishantshrivastav23, train.py from pytorch-segmentation has a --test-only option that skips training and just does eval on the dataset - is that what you are looking for?

Hi @dusty_nv ,

With --test-only we can evaluate the segmentation model on dataset after training but I am looking for doing prediction/inference and get labels and confidence on test.jpg or test.mp4 using opencv python script which will work on Jetson as well as on other system

The --test-only eval mode is doing prediction/inference, as it is not doing training.

If you want it to work on other systems as well, I recommend you write a PyTorch script for doing this as jetson.inference is only supported on Jetson at this time.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.