How to get output-tensor-meta from 2 sgie in pipeline?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) 1080 Ti
• DeepStream Version 5.0
• JetPack Version (valid for Jetson only)
• TensorRT Version 7.0.0.1
• NVIDIA GPU Driver Version (valid for GPU only) 450.80.02
• Issue Type( questions, new requirements, bugs) bugs/question

Hello!

I’m using pyds. My pipeline consist of two sgie’s:

… → tracker → pgie → sgie1 → sgie2 → …

Separately sgie1 or sgie2 can extract tensor output meta, but when I connect it, in sgie2 src probe I get tensor-output from sgie1.

3

This is not a known issue.
From your very brief informantion, I can’t tell what’s the wrong.
Could you check if you hook the probe on right src pad?

4

Yes, the probe on right src pad.
My models has the following structure.
sgie1:
0 INPUT kFLOAT input_1 3x224x224
1 OUTPUT kFLOAT output_1 44

sgie2:
0 INPUT kFLOAT input_1 3x224x224
1 OUTPUT kFLOAT output_1 128

Could equal layer names reflect to output-tensor-meta behavior?
sgie1 has batch-size=1
sgie2 has batch-size=2

6

I’ve solved the issue.

In my config, pgie is gie-unique-id=1
sgie1 has gie-unique-id=4
sgie2 has gie-unique-id=5

sgie1 has operate-on-gie-id=1
but sgie2 has operate-on-gie-id=4, when I change it to operate-on-gie-id=1

But question: Could I set for sgie2 operate-on-gie-id=4, and process-mode=1 (primary) to work on full frame, that received from sgie1, because sgie1 get crops from pgie detector? Does it will be effective or not?

7

yes, in https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/tree/master/back-to-back-detectors , two gies are primary gie