Gst-nvvideotestsrc | DeepStream

The Gst-nvvideotestsrc plugin uses CUDA to generate video test data in a variety of formats and patterns that is written directly to GPU output buffers using the video/x-raw(memory:NVMM) data type.

The plugin can also be used to input raw video and generate GPU buffers at the output. The plugin provides properties to introduce jitter in buffer latencies which can be used to simulate network conditions locally.

The Gst-nvvideotestsrc plugin is currently only supported on X86.

Outputs

This section summarizes the outputs of the Gst-nvvideotestsrc plugin.

Output
- Gst Buffer with a test pattern
- Video Format: NV12, I420, RGBA, UYVP, BGRA64_LE, RGB10A2_LE (NVMM)
Control parameters
- gpu-id
- animation-mode
- memtype
- blocksize
- pattern

Gst Properties

The following tables describes the Gst properties of the Gst-nvvideotestsrc plugin.

nvvideotestsrc plugin properties

Property	Meaning	Type and Range	Example notes
animation-mode	For animating patterns, the counter that controls the animation.	Enum “GstNvVideoTestSrcAnimationMode” Default: 0, “frames (0): frames - Frame count (1): wall-time - Wall clock time (2): running-time - Running time	animation-mode=0
blocksize	Size in bytes to read per buffer (-1 = default)	Unsigned Integer. Range: 0 - 4294967295 Default: 4096	blocksize=4096
gpu-id	ID of the GPU where the buffers are allocated	Unsigned Integer. Range: 0 - 2147483647 Default: 0	gpu-id=0
num-buffers	Number of buffers to output before sending EOS (-1 = unlimited)	Integer. Range: -1 - 2147483647 Default: -1	num-buffers=300
pattern	Type of test pattern to generate	Enum “GstNvVideoTestSrcPattern” Default: 0, “smpte (0): smpte - SMPTE color bars (1): mandelbrot - Mandelbrot set (2): gradient - Gradient	pattern=1
memtype	Type of the memory used for the allocated buffers	Enum “NvBufSurfaceMemType” Default: 2, “Device CUDA memory (1): Host/Pinned CUDA memory - host (2): Device CUDA memory - device (3): Unified CUDA memory - unified	memtype=2
do-timestamp	Apply current stream time to buffers	Boolean. Default: false	do-timestamp=true
is-live	Whether to act as a live source. Introduces latency based on FPS setting on src caps	Boolean. Default: false	is-live=true
location	File location for raw video file input. When file location is not specified, pattern gets used.	String. Default: NULL	location=test.rgba
file-loop	Loop through the input file until num-buffers (if specified) or infinitely	Boolean. Default: false	file-loop=true
max-jitter	Maximum jitter (in milliseconds) that can be applied to each buffer. Only applicable when is-live=true. A random jitter ranging up to “max-jitter” gets applied to each buffer.	Integer, ≥0 Default: 0	max-jitter=100
fixed-jitter	List of fixed jitter values (in milliseconds). The list will be looped through for the entire sequence. Only applicable when is-live=true. If “fixed-jitter” is specified “max-jitter”(if specified) is ignored.	Semicolon delimited integer array. all values ≥0. Default:	fixed-jitter=30;40;100;0;0
horizontal-speed	Horizontal speed of the SMPTE pattern in pixels per frame. Positive values move to the left; negative to the right	Integer. Default: 0	horizontal-speed=1

How to test

nvvideotestsrc is currently supported for X86 only. support with “Jetson + dGPU” is not yet enabled.

Run following reference gst-launch-1.0 pipeline on DGPU setup.

A) Default SMPTE Pattern - 720p 60Hz

gst-launch-1.0 nvvideotestsrc ! queue ! nveglglessink -e

B) Mandelbrot Pattern - 1080p 60Hz

gst-launch-1.0 nvvideotestsrc pattern=mandelbrot ! 'video/x-raw(memory:NVMM), width=1920, height=1080, framerate=60/1' ! queue ! nveglglessink -e

C) Gradient Pattern - 720p 60Hz

gst-launch-1.0 nvvideotestsrc pattern=2 ! queue ! nveglglessink -e