Embedded Intel Atom® Processors
Intel Atom® Hardware, Software, Firmware, Graphics
Announcements
This community is designed for sharing of public information. Please do not share Intel or third-party confidential information here.
1012 Discussions

FP32 TO INT8 model conversion using Calibration tool

CJOY01
Beginner
339 Views

I am facing one issue on using the  calibrate.py tool used for calibrating the FP32 model.

 

I was successfully able to convert data annotation files using the below mentioned step.

 

python3 convert_annotation.py imagenet --annotation_file /home/cyril.pj/Downloads/downloadsforint8/val.txt --labels_file /home/cyril.pj/Downloads/downloadsforint8/synset_words.txt -ss 2000 -o ~/annotations -a imagenet_calibration.pickle -m imagenet_calibration.json

 

after successful completion of the above mentioned step, i was able to generate  .json and .pickle file.

 

 

then for calibrating the model, i tried the below mentioned step,

 

 

python3 calibrate.py --config /opt/intel/openvino_2019.3.376/deployment_tools/tools/calibration_tool/configs/inception_v1.yml --definition /opt/intel/openvino_2019.3.376/deployment_tools/tools/calibration_tool/configs/definitions.yml -M /opt/intel/openvino_2019.3.376/deployment_tools/model_optimizer --tf_custom_op_config_dir ~/tf_custom_op_configs --models ~/models --source /home/cyril.pj/SSD_COPIED/rawfiles --annotations ~/annotations –cfc

 

 

but i am not able to successfully execute the calibration tool, the error i am getting is ,

 

Detected not satisfied dependencies:

networkx: installed: 2.4, required: 2.4

 

on checking the  model_optimizer/requirements_tf.txt   , the required version for  networkx is  given as  networkx>=1.11,<2.4  .

 

I have tried with couple of networkx version between 1.11 and 2.3 and installed all its dependency packages, but the same issue is coming every time.

 

Kindly look into this issue,Any help you can give would be greatly appreciated.

 

Thanks and Regards,

CYRIL P JOY

 

0 Kudos
1 Reply
CarlosAM_INTEL
Moderator
161 Views

Hello, @CJOY01​:

 

Thank you for contacting Intel Embedded Community.

 

Your consultations of this type should be addressed to the channel stated in the following website:

 

https://software.intel.com/en-us/forums/computer-vision

 

Best regards,

@Mæcenas_INTEL​.

Reply