Is it possible to add multiple class in emdx-analytics config.yaml File ? My goal is to increment tripwire count on multiple class detection

Hi, we have a YoloV7 model that was trained on some custom data to detect different kinds of artefacts of a bottle. We are using Metropolis microservices specifically we’re using the sample pipeline provided in the Quick Start guide to create our application. The input to DeepStream is a sample video of bottles moving on a conveyor belt. On configuring the analytics service to detect a single artefact with a tripwire the count of the artefact is correctly maintained. However, we wanted to maintain a count of all the artefacts that were detected.
Is it possible to include more than one class in the emdx-analytics config.yaml file so that a single tripwire can track multiple classes?
For your reference, we attached a bottle_overlay.png. Please find the attachment.

I will check it and feedback here later.

Multi-class support in the Analytics Analytics Microservices will be added in the next release. Stay tuned.

1 Like

Hi @kesong, When can we expect the next release?

The plan is early next month. Is it fine for your project?

yes, it’s fine but can you give me an idea of it will work on multi-class AND logic or OR logic?

Can you share your requirement on multi-class support in emdx-analytics? Let’s try to ensure it works in your use case.

I want to add multiple classes in emdx-analytics service and my configured trip wire count should increment when any of the mentioned classes is detected.
So is it possible in the next release?


Good to see your work.

Can you please let me know how you use a custom yolov7 model with the deepstream app available with MMJ?

NOTE: I know how to run the custom yolov7 model with normal deepstream app(without MMJ).

We have yolov8 in latest release. Please refer: DeepStream Perception — Metropolis on Jetson documentation 0.1.0 documentation

Hello @sarodeanil35 ,
Sorry for late reply due to busyness.

Using the short guide provided by NVIDIA, I first installed all the services. Next, I used deepstream configuration to switch from Peoplenet to the YOLO model. [My format is onnx].