How to split high resolution source for object detection in deepstream?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
#1

Please provide complete information as applicable to your setup.

• Hardware Platform : T4
• DeepStream Version: 5.1
• TensorRT Version: 7.2
• Issue Type: new requirements

Hi, NVs,

I need detect object in high resolution image or video with deepstream, like 4k, 8k or even higher. In now dst version, the image will be resize after decode bin and streammux plugin, and raise a problem that lots of information will be lost for object detection.

I want to split the image for multi image, for example 2x2 small image from 1 big image. So my question is how to achieve this function in deepstream? Or whether there is same example I can refer.

Thanks !

#3

May be videocrop plug can help. Let me check internally.

#4

Hello, @kesong any update? Thanks

#5

Hi @kesong ,does preprocess plugin achieve this function? Thanks.

#6

Seems ROI can meet your requirement. But need upgrade to DS 6.0.

Gst-nvdspreprocess (Alpha) — DeepStream 6.0 Release documentation (nvidia.com)

#7

OK, thanks for your reply.