Deep Stream 7.1 Gst-nvof understanding the NvDsOpticalFlowMeta results

Accelerated Computing Intelligent Video Analytics DeepStream SDK
, ,
1

Model: CTI Hadron + Orin NX - Jetpack 6.1 [L4T 36.4.0]
NV Power Mode[1]: 10W
Hardware:
Module: NVIDIA Jetson Orin NX (16GB ram)
Platform:
Distribution: Ubuntu 22.04 Jammy Jellyfish
Release: 5.15.148-tegra
Libraries:
CUDA: 12.6.68
cuDNN: Not installed
TensorRT: Not installed
VPI: 3.2.4
Vulkan: 1.3.204
OpenCV: 4.5.4 - with CUDA: NO

Hi
i have 2 questions regarding the gstreamer nvof:

  1. using python to run gstreamer pipeline on the jetson which include optical flow plugin by nvidia the Gst-nvof, i am probing the buffer of the flow vectors same as here
    i tried to test it on a solid white video using “videotestsrc pattern=white” as an input source , the flow vector x and y i am getting on this is solid no movement video is either 16 (S10.5 representation) or 0.5 (converting to floating point) for both x and y , wondering why it’s represent this way instead of 0.0 when there is not movement ?

  2. another question regarding the nvof grid-size , it works on 4x4 block, does the output flow resolution is pixel level resolution ?

Thank You

4

We need to check internally first, will be back to you soon.

The output flow vector array resolution is block level resolution.

5

@Fiona.Chen Hi , thank you , i am waiting for response

6

Hi, @anylizerexe :

We are investigating the VPI implementation of the optical flow, it may take some time. Thank you!