Deepstream Incorrect Output via NvMsgConv & NvMsgBroker

[beauy152]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.2
• TensorRT Version 8.5.2
• NVIDIA GPU Driver Version (valid for GPU only) 520.61.05
• Issue Type( questions, new requirements, bugs) Bug
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description) Nvmsg-conv, Nvmsg-Broker

I’ve been testing an application using Deepstream 6.2, nv-msg-conv & nv-msg broker. the pipeline is setup as:

pgie → tracker → sgie_1 → sgie_2 → sgie_3 → sgie_4 → nvvideo convert → nvdsosd → tee (nvmsg-conv → nvmsg-broker) | (nveglglessink)

i’ve noticed the following things (all with msg2p-newapi=1 & multiple-payloads=1 set on nvmsg-conv):

• when using ‘payload-type=0’, there is only ever a single object attached in the payoad (even though multiple-payloads is set). Additionally, any object will also never have any of the secondary classifier values attached from the sgie’s.
• when using payload-type=1, there will be multiple objects attached, and the secondary classifier values will be attached. Although, only objects which have secondary classifier data attached will get their confidence values attached. Confidence values are also not attached for any of the secondary classifier values which were attached.

I could not find any indication that these are known issues yet or not.

[zhy2022]

@beauy152

I could not find any indication that these are known issues yet or not.

I recommend you reading source code of plugin msg-conv and msg-broker provided in deepstream devel docker image.

IMHO, there’s some parameters and logic hard to understand in there two plugins.

[fanzh]

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

@beauy152

Although, only objects which have secondary classifier data attached will get their confidence values attached. Confidence values are also not attached for any of the secondary classifier values which were attached.

nvmsgconv is opensource, please refer to opt\nvidia\deepstream\deepstream-6.2\sources\libs\nvmsgconv\nvmsgconv.cpp, you can add log to check.