Mmdetection object detection model on jetson agx xavier

AI & Data Science Deep Learning (Training & Inference) TensorRT
, ,
1

Description

Hey there, I am currently trying to get a object detection & instance segmentation algorithm from MMDetection running on my Jetson AGX Xavier. For setting up the environment, I am following the procedure here: Prerequisites — MMDetection 2.20.0 documentation

I have created a new conda environment and I am trying to install pytorch on it. (conda install pytorch torchvision cudatoolkit=10.2 -c pytorch) but this keeps throwing me the following error:

PackagesNotFoundError: The following packages are not available from current channels:

  • torchvision=0.4.2
  • cudatoolkit=9.2
  • pytorch=1.3.1

I am able to install pytorch using pip (pip3 install torch torchvision). However, I cannot access cuda from it as is throws False if I type import torch → torch.cuda.is_available().

Has anyone encountered these issues on a Jetson and do you know a way to solve this issue?

Furthermore, from searching for this on the Web I am doubting if MMDetection is really suited for the Jetson. I am reading a lot about using ONNX to convert the model to TensorRT. I would be very happy if someone could give me some advice on how to proceed with this!

Thanks a lot and best regards,

Joël

Environment

TensorRT Version: 8.2.1
CUDA Version: 10.2

2

Hi,

This looks like a Jetson issue. Please refer to the below samples in case useful.

For any further assistance, we will move this post to to Jetson related forum.

Thanks!