DeepStream Function Expansion

[2536758546]

Please provide complete information as applicable to your setup.

• Hardware Platform == Jetson
• DeepStream Version == 6.3
• JetPack Version == 35.6.0
• TensorRT Version == 8.5.2.2
• NVIDIA GPU Driver Version == 11.4.315
• Issue Type == new requirements

I have just come into contact with the deepstream application and want to develop some custom functions based on the Deepstream-test5 sample program, and my code runs based on deepstream-test5-app-c my_config.txt. I currently have the following requirements:

1. When running deepstream, switch to a custom detection model. I saw that test5 provides OTA, but it requires that the model detection category before and after the switch be consistent with the push size. Excuse me, what should I do?
2. When running deepstream, modify the detection confidence threshold and it will take effect in real time.

I would be very grateful to you if you could provide me with a general development idea or plan.

[yuweiw]

@2536758546

When running deepstream, switch to a custom detection model. I saw that test5 provides OTA, but it requires that the model detection category before and after the switch be consistent with the push size. Excuse me, what should I do?

This is to ensure that the buffers of the already running deepstream-app are reused. We currently do not support adding different models.

@2536758546

When running deepstream, modify the detection confidence threshold and it will take effect in real time.

This part is open-source. You can try customizing it to meet your requirements.