Setup Re-ID model issue

I am trying to follow steps in sources/tracker_NvDeepSORT/README to setup the Re-ID model. But I am getting the following two errors. This is after downloading the mars-small128.pb file and placing it under sources/tracker_NvDeepSORT/. Is this due to imcompatible tf version?

My environment:
•Hardware Platform: RTX 3070 Mobile
•DeepStream Version: 6.2-devel container
•TensorRT Version:
•NVIDIA GPU Driver Version (valid for GPU only): 525.105.17

In second step it specifies to install tensorflow and pyyaml by running $ pip3 install tensorflow-gpu PyYAML ,but I get the error

error: subprocess-exited-with-error
  × python bdist_wheel did not run successfully.
  │ exit code: 1
  ╰─> [18 lines of output]
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "/tmp/pip-install-ov42g02p/tensorflow-gpu_6e3590eaee7244b5ae23b0854e280c4c/", line 37, in <module>
          raise Exception(TF_REMOVAL_WARNING)
      The "tensorflow-gpu" package has been removed!
      Please install "tensorflow" instead.
      Other than the name, the two packages have been identical
      since TensorFlow 2.1, or roughly since Sep 2019. For more
      information, see:
      [end of output]

Then I tried $ pip3 install tensorflow PyYAML and it installed tensorflow correctly. Third step specifies just to run $ python3 , but I get the following error

root@jorge-ASUS-TUF-Dash-F15-FX517ZR-FX517ZR:/opt/nvidia/deepstream/deepstream-6.2/sources/tracker_NvDeepSORT# python3
2023-05-03 20:22:18.963157: I tensorflow/core/util/] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2023-05-03 20:22:18.986814: I tensorflow/core/platform/] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
Traceback (most recent call last):
  File "", line 19, in <module>
    import uff
  File "/usr/lib/python3.8/dist-packages/uff/", line 48, in <module>
    from uff import converters, model  # noqa
  File "/usr/lib/python3.8/dist-packages/uff/model/", line 48, in <module>
    from . import uff_pb2 as uff_pb  # noqa
  File "/usr/lib/python3.8/dist-packages/uff/model/", line 33, in <module>
  File "/usr/local/lib/python3.8/dist-packages/google/protobuf/", line 796, in __new__
TypeError: Descriptors cannot not be created directly.
If this call came from a file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.
If you cannot immediately regenerate your protos, some other possible workarounds are:
 1. Downgrade the protobuf package to 3.20.x or lower.
 2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).

More information:

Thanks for reporting it. I will check internally and feedback.

tensorflow-gpu has been removed recently, so the command is changed into

pip3 install tensorflow==2.11.0 PyYAML

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