Why does nvmultistreamtiler change number of detected objects stored in NvDsBatchMeta seemingly randomly?

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU): GeForce RTX 3090
• DeepStream Version: 5.1
• TensorRT Version: 7.2
• NVIDIA GPU Driver Version (valid for GPU only): 460.90
• Issue Type( questions, new requirements, bugs): question

I updated deepstream_test1_app.c (17.9 KB) to the following pipeline:

cam1_source \
              --> nvstreammux --> nvinfer --> queue --> nvmultistreamtiler --> nvvideoconvert --> nvdsosd --> nveglglessink
cam2_source /

I noticed that sometimes, the number of objects detected in a batch change after the batch meta passed through nvmultistreamtiler plugin, is this expected behaviour?

Running...
** Message: 15:40:02.496: test_probe TILER SINK: num_objects: 0
** Message: 15:40:02.497: test_probe TILER SRC: num_objects: 0
** Message: 15:40:02.560: test_probe TILER SINK: num_objects: 1
** Message: 15:40:02.561: test_probe TILER SRC: num_objects: 1
** Message: 15:40:02.667: test_probe TILER SINK: num_objects: 1
** Message: 15:40:02.667: test_probe TILER SRC: num_objects: 2

** (deepstream-test1-app:14642): ERROR **: 15:40:02.667: osd_sink_pad_buffer_probe MISS-MATCH!!! meta_counts = 1, counts = 2
Trace/breakpoint trap (core dumped)

Sorry for the late response, is this still an issue to support?

@kayccc , thanks for the response, it’s still an issue

Hi any update on this?