• Hardware Platform (Jetson / GPU)
Jetson
• DeepStream Version
deepstream-app version 6.0.1
DeepStreamSDK 6.0.1
• JetPack Version (valid for Jetson only)
JetPack 4.6, L4T 32.6.1
• TensorRT Version
TensorRT 8.0.1
• NVIDIA GPU Driver Version (valid for GPU only)
Non GPU version
• Issue Type( questions, new requirements, bugs)
Question
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
Run the default deepstream-nvdsanalytics-test using the provided configuration file, and capture the output in a log file.
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)
Requirements to run the default deepstream-nvdsanalytics-test app.
The question I have is that regardless of the interval I choose in the main configuration file for the application, under the [primary gie] field I choose, the frames outputted from the inference seem to not vary.
E.g. when I run the deepstream-nvdsanalytics-test app with an interval of 0 and a batch-size of 1, I can see that no frames are skipped in the output, so far so good, but then I chose a 20 interval with a 1 batch-size the output remains the same, meaning that I can see frames outputted counting from frame number 0 to N (the same as if the inference was not skipping any frames), but regarding GPU usage it seems to respect the interval configuration.
Using for reference the definitions provided in deepstream-app reference application. Interval is Number of consecutive batches to skip for inference. and batch-size is The number of frames(P.GIE)/objects(S.GIE) to be inferred together in a batch. From both of these definitions I was expecting that if I run the interval 20, batch-size 1 config for the app, the frame 0 would be inferred, then 20 frames would be skipped, then frame 21 would be inferred and so on…
Am I misunderstanding something here? Any pointers would be greatly appreciated, thanks beforehand.
Configurations used are provided here file_configuration.tar.xz (8.5 MB)
Screenshots from both apps runs are provided here, followed by jtop screenshots to display GPU usage differences:
Logs from interval 0 and batch-size 0:
interval0.txt (14.0 KB)
Logs from interval 20 and batch-size 0:
interval20.txt (18.4 KB)