Is there a way to guarantee that frame(t+1) is not processed until frame(t) is processed by the all plugins in the pipeline?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.0
• JetPack Version (valid for Jetson only)
• TensorRT Version 8.2.3.0
• NVIDIA GPU Driver Version (valid for GPU only) 470.141.03
• Issue Type( questions, new requirements, bugs) Question

Hi,

I’m running a pipeline that looks like the following:

nvinfer → custom_plugin_1 → nvinfer → custom_plugin_2

While debugging, I noticed that frame(1) is processed by custom_plugin_1 before frame(0) gets to custom_plugin_2.

Is there a way to guarantee that frame(t+1) is not processed until frame(t) is processed by all plugins in the pipeline?

Thanks!

3

you can add probe function A on custom_plugin_2’src. when entering A once, you can input one frame to nvifner.

4

Okay, then we need to craft a synchronization mechanism, we removed nvinfer for easy replication.

src > queue > nvstreammux > nvconv > queue > p1 > queue > p2 > sink
, where p1, p2 are the two custom plugins.

Case 1 Muxer Configs:

  • One Source only connected to nvstreammux
  • nvstreammux Batch Size = 30

Upon inspecting the very first NvDsBatchMeta that enters p1, we find that it contains multiple frames (2 to 5 frames) for the same source. This is not desired as we only need 1 frame per source per batch.

Case 2 Muxer Configs:

  • One Source only connected to nvstreammux
  • nvstreammux Batch Size = 1

Upon inspecting the very first NvDsBatchMeta that enters p1, we find that it contains 1 frame per source per batch. This allows p2 to work on it as well and unblock p1 to process the next frame for that source (since it’s blocked because p2 is yet to process the first frame)

So simply put, Is there a way to set a batch size that is higher than current muxer sources and force 1 frame per source in NvDsBatchMeta?

5

about how to set nvstreammux’s batch-size, please refer to this topic nvstreammux-batch-size