Pipeline not working as intended with slow plugin

I am using deepstream 3, with pipeline adapted from the old yolo plugin in https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps

The pipeline is as follows
add uridecodebin source
gst_link_many(nvstreammux, nvvidconv, queue, yolo_plugin, appsink, NULL);

To simulate a slow plugin, I sleep for 5 seconds in yolo_plugin whenever it receives a frame upstream. Despite having a queue, it does not give me the latest frame, but it gives me the next frame from the camera stream.

Can anyone help me to understand why is there no dropping of frames despite having a queue?

If you are using Nvidia Jetson nano, please make sure your power connection. If you are use micro usb port in you power connection then yolo will be work very slowly. That’s better if you are use a power Supply 4 Amps at 5 Volts through the Barrel Jack connector in your Jetson device. Please check this article, you will know about more details. https://www.jetsonhacks.com/2019/04/10/jetson-nano-use-more-power/

I am not using Nvidia Jetson Nano, I am using a T4, but the gpu I am using is irrelevant. I want to understand more about my gstreamer pipeline, specifically, what happens to the streams when one element is slow, and why is the queue element not working

Hi, can I understand why is rtspsrc in uridecodebin not running constantly? Instead, it only runs every 5 seconds, since yolo element is sleeping for 5 seconds.

Hi, can any nvidia moderator respond?

Please go to DS 4.0.1 as early as possible.

Check with qos = 1 and drop-on-latency = TRUE for rtspsrc