The most complete jetson in history uses jetpack4.6.1 to burn, configure and configure pytorch and tensorrt in the virtual environment (3)

3.1 Jetson uses gstreamer to control the csi camera

Of course, you can also directly plug in the usb camera, which can be used in the same way as a laptop camera

(44 messages) Zero-based entry to Jetson Nano——calling CSI and USB camera through OpenCV_jetson nano calling usb camera_DaDa️'s blog-CSDN blog The camera can be configured like this, but you will encounter a bunch of errors :

1. After using the camera once, it cannot be turned on the second time: you can restart it or see the link below

(44 messages) Error generated. /dvs/git/dirty/git-master_linux/multimedia/nvgstreamer/gst-nvarguscamera/gstnvargus

2. Various gstremer warnings

[ WARN:0] global /home/ubuntu/build_opencv/opencv/modules/videoio/src/cap_gstreamer.cpp (1390) setProperty OpenCV | GStreamer warning: 

Please add a parameter

cap = VideoCapture(0, cv.CAP_V4L2)

it's done

3.2 The use of tensorrt

There are two ways here

sudo cp -r /usr/lib/python3.8/dist-packages/tensorrt* /home/nx/miniforge-pypy3/envs/Torch8/lib/python3.8/site-packages/

It is also possible to replace j or cp -r with ln for soft connection

The blogger's model uses leaky-relu, which is not supported in tensorrt8.0, so the blogger will configure it again in jetpack5.1 to use tensorrt8.5