How to specify the detection area in the screen?

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson)
**• DeepStream Version5.1
• JetPack Version (4.5)
**• TensorRT Version7.x
I successfully run the official sample program, but I want to make the following modifications to sample source8_1080p_dec_infer-resnet_tracker_tiled_display_fp16_nano.txt

I only want to detect objects in the middle area, not the whole picture. For example, the ROI range is 600x600 in the middle of the picture, not 1920x1080.

How can I do this in example source8_1080p_dec_infer-resnet_tracker_tiled_display_fp16_nano.txt?

Currently we don’t support this, but you can filter out the objects via roi-top-offset and roi-bottom-offset , refer Gst-nvinfer — DeepStream 5.1 Release documentation

Can’t I use the nvdsanalytics in DP5.1? I just saw it, the ROI filtering.

Yeah, you can use that pulgin to do the analytics, but it cannot specify which ROI should be put to the detector. In general this plugin should be a downstream to the detector

1 Like