Tools and procedures for general inspections or debugging of the detection pipeline in deepstream 5.0

Hi,
I’m wondering if there is any tools or procedures we can use for general inspections or debugging of the detection pipeline in deepstream 5.0?

For instance, if I found out that objectDetector-Yolo app failed to detect objects in certain frames through a usb-cam, how can I intercept intermediate results before/during/after each steps including receiving raw image input from camera, preprocessing, inference and post-processing?

My settings are the following:
Jetson Xavier
DeepStream 5.0
JetPack 4.4
TensorRT 7.1.3
NVIDIA GPU Driver Version 10.2

Hi,

Please check below sample to get the buffer and save the data:

/opt/nvidia/deepstream/deepstream-5.0/sources/gst-plugins/gst-dsexample

Thanks.

Thanks your response! It would help solve the specific example issue I raised. However, for a more general question (this could be a long shot), is there any tool available for a DS pipeline to run a self-inspection?

Hi,

You will need to save the data to check the intermediate buffer value.
So please use the example shared above to achieve this.

If you want to check bug level issue, please see below document for information:
https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_FAQ.html#how-can-i-run-the-deepstream-sample-application-in-debug-mode

Thanks.