DeepStream-app test 5 build custom ReID nodel engine but can not deserialized it

Please provide complete information as applicable to your setup.

• Hardware Platform GPU
• DeepStream Version 6.4
• TensorRT Version 8.6
• NVIDIA GPU Driver Version (valid for GPU only) 535.161.07
• Issue Type( questions, new requirements, bugs) bug

Description

I try to add custom ReID model in [tracker] pipeline for deepstream-app-test5. So I have onnx model, add it to tracker config file:

# [Paths and Names]
  onnxFile: "/media/hhd/ntn/ntn_21/ntn/DeepStream-Yolo/weights/osnet_ain_x1_0_msmt17.onnx"
  modelEngineFile: "/media/hhd/ntn/ntn_21/ntn/DeepStream-Yolo/weights/osnet_ain_x1_0_msmt17.onnx_b100_gpu0_fp16.engine" # Engine file path

For 1st run, when engine model was not generated, deepstream build engine and ran app successfully, but for 2nd run it could not deserialized built engine file:

neonick@neonick:/media/hhd/ntn/ntn_21/ntn/DeepStream-Yolo$ deepstream-test5/deepstream-test5-app -c rtsp_record_tracking_msg_log_creid_config.txt -t 1
gstnvtracker: Loading low-level lib at /opt/nvidia/deepstream/deepstream-6.4/lib/libnvds_nvmultiobjecttracker.so
[NvMultiObjectTracker] Loading TRT Engine for tracker ReID...
ERROR: [TRT]: 3: getPluginCreator could not find plugin: InstanceNormalization_TRT version: 1
ERROR: [TRT]: 1: [pluginV2Runner.cpp::load::303] Error Code 1: Serialization (Serialization assertion creator failed.Cannot deserialize plugin since corresponding IPluginCreator not found in Plugin Registry)
~~ CLOG[src/modules/ReID/ReID.cpp, loadTRTEngine() @line 645]: Engine build failed
[NvMultiObjectTracker] Load engine failed. Create engine again.

Can you share the code and configuration files so I can reproduce the problem?

Are you using c or python? If you are using python, here is a similar topic.

No, Im not using python. Just edit follow lines in config_tracker_NvDeepSORT.yml file

  reidFeatureSize: 512    # Size of reid feature
...
# [Paths and Names]
  onnxFile: "/media/hhd/ntn/ntn_21/ntn/DeepStream-Yolo/weights/osnet_ain_x1_0_msmt17.onnx"
  modelEngineFile: "/media/hhd/ntn/ntn_21/ntn/DeepStream-Yolo/weights/osnet_ain_x1_0_msmt17.onnx_b100_gpu0_fp16.engine" # Engine file path

Is this a model you trained yourself?

No
“osnet_ain_x1_0_msmt17.pt”: “Google Drive - Quota exceeded”,

from this repo: https://github.com/mikel-brostrom/yolo_tracking/blob/master/boxmot/appearance

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

sorry for the long delay.
The configuration file of this tracker can only support resnet50_market1501.etlt and does not support any model.

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