- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello, I work for Siemens. We have brought Intel® Movidius™ Neural Compute SDK toolkit for Tensor flow Models. Have been working for Attention OCR Model(https://github.com/kspook/Attention-OCR-1 ). After Model training, am struggling to convert the model results into 4shv file (getting lots of error- now struggling with this error - "ValueError: Node 'map_1/while/foldr/while/cond/add/Switch' expects to be colocated with unknown node 'map_1/while/foldr/while/Identity_1' Makefile:75: recipe for target 'compile' failed ") I just want to know whether Kit supports the AOCR model. Have attached the checkpoints and frozen graph including the input and output node as well.
Please help me to convert the required file to 4shv format.
Thanks,
Vasuki
- Tags:
- Tensorflow
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi VHebb1,
I've looked at the model, and it looks like it would not be compatible on the NCSDK since this AOCR model is an LSTM on top of a CNN. Also, the NCSDK is now unsupported. As previously stated by David, I think you have more chances of your model working on OpenVINO. Since part of the model is in Keras, you will need to convert to Tensorflow, and then make sure that it falls into the supported topologies list to be able to convert and run on OpenVINO.
I hope this information is helpful. Please let me know if you have any further questions.
Best Regards,
Sahira
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi VHebb1,
Thanks for reaching out.
The Intel® Movidius™ NCSDK is no longer being maintained. We recommend you using the OpenVINO™ toolkit, you can review this guide to transition from the Intel® Movidius™ Neural Compute SDK to Intel® Distribution of OpenVINO™ toolkit.
If you have additional questions, let us know.
Best regards,
David C.
Intel Customer Support Technician
A Contingent Worker at Intel
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Answer is not at all relevant to the question that i have asked. Its a pure technical question. Can you pass this question directly to technical people ?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi VHebb1,
I've looked at the model, and it looks like it would not be compatible on the NCSDK since this AOCR model is an LSTM on top of a CNN. Also, the NCSDK is now unsupported. As previously stated by David, I think you have more chances of your model working on OpenVINO. Since part of the model is in Keras, you will need to convert to Tensorflow, and then make sure that it falls into the supported topologies list to be able to convert and run on OpenVINO.
I hope this information is helpful. Please let me know if you have any further questions.
Best Regards,
Sahira
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hello,
Thank you so much for the answer. It was indeed helpfull.
Even I tried TensorFlow Slim Network as well, couldn't achieve.
Can you please let me know if this does support or not.
Guidelines for Tensorflow - https://movidius.github.io/ncsdk/tensorflow.html
https://github.com/tensorflow/models/tree/master/research/attention_ocr
Thanks,
Vasuki
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi Vasuki,
To convert a Tensorflow Slim Model, you can use the Intel® Distribution of OpenVINO™ toolkit. Instructions for converting your Tensorflow Slim Model are here.
It will not work on the NCSDK.
I hope this information is helpful.
Best Regards,
Sahira
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page