get an error loading a model for auto-annotation

[test4fest]

I am trying to use Faster-RCNN model from model zoo of auto annotation from the following link:

https://github.com/opencv/cvat/tree/develop/utils/open_model_zoo/faster_rcnn_inception_v2_coco

I follow the documentation and convert tensorflow model of faster- RCNN to xml and bin files with OpenVini 2019- R3, but when I ran CVAT with OpenVino 2018-R5.
I get this error:

Checking request has returned the "failed" status. Message: Exception: Model was not properly created/updated. Test failed: Error reading network: cannot parse future versions: 6

Does it mean the conversion is not correct, or using different version of OpenVino, or something else?

[benhoff]

Different version of OpenVino. You should be able to use OpenVino - R5 with the CVAT installation. Try reinstalling the OpenVino component with R5.

[test4fest]

The problem remains unsolved. I can run auto-annotation with person-vehicle pre-trained model but not for faster-RCNN .

[benhoff]

If you're a Python developer, can you run the following script and post the traceback?

https://github.com/opencv/cvat/tree/develop/utils/auto_annotation

Otherwise I'll try to reproduce tomorrow.

[benhoff]

I tried with my local system, but was not able to reproduce. I would try cleaning the docker image. You've probably got a stale OpenVINO installation.

[test4fest]

If you're a Python developer, can you run the following script and post the traceback?

https://github.com/opencv/cvat/tree/develop/utils/auto_annotation

Otherwise I'll try to reproduce tomorrow.

Thanks for your reply.
can you share xml and bin file for pre-trained Faster-RCNN network with us?

I got the following error:

[setupvars.sh] OpenVINO environment initialized
Traceback (most recent call last):
File "/home/django/utils/auto_annotation/run_model.py", line 136, in
main()
File "/home/django/utils/auto_annotation/run_model.py", line 100, in main
restricted=restricted)
File "/home/django/utils/auto_annotation/../../cvat/apps/auto_annotation/inference.py", line 125, in run_inference_engine_annotation
model = ModelLoader(model=model_file, weights=weights_file)
File "/home/django/utils/auto_annotation/../../cvat/apps/auto_annotation/model_loader.py", line 24, in init
network = make_network(self._model, self._weights)
File "/home/django/utils/auto_annotation/../../cvat/apps/auto_annotation/inference_engine.py", line 40, in make_network
return IENetwork(model = model, weights = weights)
File "ie_api.pyx", line 266, in openvino.inference_engine.ie_api.IENetwork.cinit
RuntimeError: Error reading network: cannot parse future versions: 6

[benhoff]

That's frustrating! Sorry about that.

Related: openvinotoolkit/open_model_zoo#71

Sorry, the models were uploaded a little bit earlier than the upcoming release itself. Once it's out (hopefully, this week) you should be able to run the model. For now, you're better off using the 2018 branch.

Here's the line of code that is throwing the error.

Potential that they've released the models without the updated code? Or that OpenVINO messed up its versioning. But I don't know that that makes sense given that you hand converted the bin/xml files. Maybe try converting the faster-rcnn bin/xml with the 2019 R3 candidate (or an earlier version) and then using the latest version of OpenVINO in your CVAT instance. The above linked issue mentions converting with the 2018 version, which I believe is what I converted with.

I was going to post my bin and xml files to github, but it's over the 10 MB limit on Github, (zipped files together are 200 MB). If you've got a better site to get them to you, let me know and I'll put them up.

[benhoff]

If you use the 2018 version, it looks like it will convert to version 3. See here: https://github.com/opencv/dldt/blob/669bee86e580cbbc8ef40b440ab195ba2cbf5142/inference-engine/src/inference_engine/ie_cnn_net_reader_impl.cpp#L144

[benhoff]

@test4fest did you ever get this fixed? What was the resolution?

[savan77]

@benhoff Using this (https://github.com/opencv/dldt/releases/tag/2018_R5) version solved error for me.

[anujonthemove]

Different version of OpenVino. You should be able to use OpenVino - R5 with the CVAT installation. Try reinstalling the OpenVino component with R5.

I am using OpenVINO version 2019.R3 (l_openvino_toolkit_p_2019.3.376), got error on console: Checking request has returned the "failed" status. Message: Exception: Model was not properly created/updated. Test failed: module 'cvat.apps.auto_annotation.inference' has no attribute 'run_inference_engine_annotation', I then tried out https://github.com/opencv/cvat/tree/develop/utils/auto_annotation but on similar error while running it on terminal ImportError: cannot import name 'run_inference_engine_annotation' .

[azhavoro]

@anujonthemove Hi, #1826 should fix this issue.