CUDA Accelerated GStreamer Camera Undistort - Thor

CUDA Accelerated GStreamer Camera Undistort
Undistort Basics
Getting Started
Evaluating the Undistort Getting the code Building the Undistort
User Guide
Camera Calibration
Examples
Using Gstd Other pipelines
Performance
Thor Orin Xavier Nano
Contact Us

The performance of the CUDA undistort element depends mainly on the input image resolution.

The following sections show the measurements of the cuda-undistort (FPS and Processing time) for the following resolutions.

HD=1920x1080
4K=3840x2160

Platform Setup

The performance measurements were done with the AGX Orin at MAXN mode, which can be activated with

sudo nvpmodel -m 0

The following sections show a comparison between using the platform at a maximum frequency (With jetson_clocks) and in base mode. This mode can be set as follows:

sudo /usr/bin/jetson_clocks

Pipeline structure

The general structure of the pipeline used for the latency measurements is shown below, for the Fisheye model.

GST_DEBUG='GST_TRACER:7' GST_TRACERS='proctime' gst-launch-1.0 filesrc location=brownconrady_distortion.mp4 ! qtdemux ! h264parse ! avdec_h264 ! nvvidconv ! 'video/x-raw(memory:NVMM),height=3840,width=2160' ! rrcudaundistort distortion-model=brown-conrady camera-matrix="{\"fx\":9.5211633874478218e+02, \"fy\":9.4946222068253201e+02, \"cx\":6.8041416457132573e+02, \"cy\":3.1446117133659988e+02}" distortion-parameters="{\"k1\":3.8939572818197948e-01, \"k2\":-5.5685725182648649e-01, \"k3\":2.3785352925072494e+00, \"k4\":-1.2037220289124213e+00}" ! perf ! fakesink -v

Results

The following table shows the results for the following platforms.

• Thor with JP 7.2

Table 1: Performance results for CUDA Undistort.

Input	CPU Load(%)	RAM(%)	GPU Load(%)	Processing time (ms)	Framerate (FPS)
HD	6.920	0.291	2.09	0.756	166.454
4K	5.242	0.446	6.00	1.851	114.919