Object Detection with General Compact S1 does not show anything on WebStream

[dmitriyteteruk]

Hi! Default config works fine on mine VAIDK. I followed tutorial and trained my model Object Detection using General Compact S1 domain because General Compact does not work with VAIDK anymore. Then I exported it and uploaded to Blob. But when I change ModelZipUrl inside of Module Identity Twin for AIVisionDevKitGetStartedModule I cannot see anything recognized (detected) on WebStream. Reported properties of Module Twins are correct.
How to check if model is running? Can someone share experience of CustomVision.ai with Object Detection using General Compact S1 domain option?
I can see that messages come to IoTHub in metrics, but I cannot read them using Azure IoT Explorer. Can somebody from Project Team check if Object Detection using General Compact S1 domain option works at all?

[v-iskha]

@tsampige - Similar issue reported on #337.

[tsampige]

What version of the azureeyemodule are you running on the device? We had addressed a bug related to running specific CV models on azureeyemodule as part of 2103-2 release. if you are no lower version of the module, then you can upgrade it by going to the azure percept studio, devices, goto vision and you'll be presented with upgrade button.

[tsampige]

one thing i forgot to ask is whether this request is pertaining to model deploying on VAIDK or Azure Percept Devkit?

[dmitriyteteruk]

one thing i forgot to ask is whether this request is pertaining to model deploying on VAIDK or Azure Percept Devkit?

This is all about VAIDK

[1297rohit]

Is there any update on this ? I also facing the same issues. The custom trained General Compact S1 model which was trained on custom vision is not working on VAIDK.

[1297rohit]

When we update the model url then this value "RtspDataUrl": null is updated and after that "VideoAnalyticsEnabled": false After this the model is not able to predict anything. Does anyone has any idea regarding this. Why does this happen after updating the model with custom vision model instead of the default mobilenet-SSD model.

[MikeSpark]

Same issue here, nothing is detected (bounding box not showing)