Processing of delayed inference data

Please provide complete information as applicable to your setup.

**• Hardware Platform (Jetson / GPU)**Jetsonnano
• DeepStream Version5.1
**• JetPack Version (valid for Jetson only)**4.4.1

• Requirement details. We are using Nano, and want to process pose estimation at around 10 FPS. Whereas, we doubt that Nano can support that. In this case, is there a possibility to leverage Buffers in either Gstreamer or DeepStream to buffer the data so that we don’t loss any captured frames through live CSI camera?

@karunakar.r can you provide details of your cameras?

@karunakar.r you can add and manage Gstreamer buffers in DeepStream through the Gst-nvstreammux plugin properties in the config file. You can refer the documentation for more details.

@bgiddwani We are currently using e-CAM24_CUNX global shutter camera.

@amlan.hackassin if I am not wrong, mux be used to consume multiple sources. My requirement is different, I want to have buffer so that delay in inference won’t have any impact on capturing of video.

@karunakar.r you can have a single source too. No issues. DeepStream relies on nvvstreammux data structure to attach the necessary metadata for the buffers. To have a clear idea, refer this documentation.