Suggest you to search in google. Maybe something wrong in the command line.
So I managed to run docker and install dependencies. After I generate engine I tried inference and it is correct.
But when I run this newly generated engine in my deepstream application error still persist.
Does this mean I’m not doing preprocessing correctly?
Thanks for the info.
Maybe it is the reason. You can try to narrow down by using tao_deploy/nvidia_tao_deploy/cv/lprnet/scripts/inference.py at main · NVIDIA/tao_deploy · GitHub…
After analyzing tao_deploy/nvidia_tao_deploy/cv/lprnet/dataloader.py I see that the image is loaded as RGB, resized, normalized and transposed to CHW. If I understood correctly this is also enabled in sgie config like this:
network-input-order: 0
net-scale-factor: 0.00392156862745098
model-color-format: 0
However, the model still makes the same mistake, so is there anything else I can do?
From above steps, I suggest you to try to follow steps in GitHub - NVIDIA-AI-IOT/tao-toolkit-triton-apps: Sample app code for deploying TAO Toolkit trained models to Triton and run LPRNet to check if there is fine.
Its post-processing is in tao-toolkit-triton-apps/tao_triton/python/postprocessing/lprnet_postprocessor.py at main · NVIDIA-AI-IOT/tao-toolkit-triton-apps · GitHub.
The official app to run lpr with deepstream is to use GitHub - NVIDIA-AI-IOT/deepstream_lpr_app: Sample app code for LPR deployment on DeepStream. You can also leverage it.
Can you confirm GitHub - NVIDIA-AI-IOT/tao-toolkit-triton-apps: Sample app code for deploying TAO Toolkit trained models to Triton can work on Jetson?
I tried but I keep getting error.
Upon better look at prerequisites I found this issue where they say nvidia-driver cannot be installed on Jetson.
There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks
The error comes form tao-toolkit-triton-apps/docker/Dockerfile at main · NVIDIA-AI-IOT/tao-toolkit-triton-apps · GitHub. This github can run on dgpu only.