How to create cascade pipeline in DS

[andhover]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU)
• DeepStream Version
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs)
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

I’ve seen the example with back-to-back detector (deepstream_reference_apps/README.md at master · NVIDIA-AI-IOT/deepstream_reference_apps · GitHub)
But I need smth different.
I want to introduce probe after first nvinfer that process boxes and find one box with the max confidence after that I want to pass the coordinates to the next nvinfer that will crop the image on box coordinates and infer on cropped image. The reason I need to crop is that on the second inference will be small objects and if I’ll do inference on the whole frame the objects will be small and the model will not be able find them.

[Fiona.Chen]

@andhover

I want to introduce probe after first nvinfer that process boxes and find one box with the max confidence after that I want to pass the coordinates to the next nvinfer that will crop the image on box coordinates and infer on cropped image.

Yes, you can add such probe function and do the thing by removing the object meta with low confidence. The API nvds_remove_obj_meta_from_frame can help you to remove the objects you don’t want. NVIDIA DeepStream SDK API Reference: Metadata Structures | NVIDIA Docs
The SGIE will do the crop inside nvinfer, you don’t need to do anything. The back-to-back sample has shown you the SGIE(nvinfer with LPD model) only do inference on the license plate objects.