Deepstream-test5 delivers the message that the target is not recognized with a negative confidence and a coordinate of 0

Please provide complete information as applicable to your setup.

**• Hardware Platform (Jetson / GPU) jetson agx orin 64GB
**• DeepStream Version 7.0
**• JetPack Version (valid for Jetson only) 6.0 GA
**• TensorRT Version 8.6.2.3
• NVIDIA GPU Driver Version (valid for GPU only)
**• Issue Type( questions, new requirements, bugs) bugs

hi
When I used the deepstream-test5 example, I added a rtsp stream through the rest api and sent messages without a target, with a negative score and a coordinate of 0 ,

XADD “33001” 1722581306906-0 “message” "{
"version" : "1.0",
"sensorId" : "303239333436003d209d8214ded00000",
"@timestamp" : "2024-08-02T06:48:26.904Z",
"frameId" : 16979,
"frameWidth" : 960,
"frameHeight" : 544,
"objects" : [
{
"trackingId" : 5,
"name" : "Person",
"score" : -0.10000000149011612,
"left" : 0,
"top" : 0,
"width" : 0,
"height" : 0
}
],

Below is my config file
configs.zip (7.8 KB)

I am checking.

you did not set pre-cluster-threshold for [class-attrs-0] or other classid. please refer to the configuration file in deepstram-test1.

I have reconfigured the file, but found that there are still negative values. These are my infer configuration file and label file, please help me see what is wrong with my configuration
config_infer_primary_RN34_PN26_960x544_dla0_orin_unprune_agx.txt (1.1 KB)
labels_peoplenet.txt (16 Bytes)

  1. noticing score, frameWidth can’t be found in nvmsgconv, did you modify the nvmsgconv code? where is the score from?
  2. nvinfer plugin is opensource. it used to do inference and add object meta. if setting [class-attrs-all], in theory, the inference with the negative score will be skipped. you can add log in DetectPostprocessor::parseBoundingBox of opt\nvidia\deepstream\deepstream\sources\libs\nvdsinferserver\infer_postprocess.cpp to check if the negative value is dumped. especially you need to rebuild and replace the /opt/nvidia/deepstream/deepstream/lib/llibnvds_infer.so.
    you can add log in attach_metadata_detector of opt\nvidia\deepstream\deepstream\sources\gst-plugins\gst-nvinfer\gstnvinfer_meta_utils.cpp to check if there is negative obj_meta->confidence. especially you need to rebuild and replace the /opt/nvidia/deepstream/deepstream/lib/gst-plugins/libnvdsgst_infer.so.

Please help me check if there is anything wrong with my configuration file

I did not find any abnormality. To simplify the pipeline, you can test your model in deeptream-test1. please refer to my last comment. you can add log to check if there are object meta with the negative score and abnormal coordinates.

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.