Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Lejboelle__Aske
Beginner
161 Views

OpenVINO model optimization - tensorflow not found.

Jump to solution

Hi,

I'm trying to convert frozen tensorflow models to IR models using the OpenVINO model optimizer. I run the following command:

sudo python3 mo.py --input_model /home/asrl/openvino_models/mobilenet_v1_0.5_160_frozen.pb

but get the following error:

Detected not satisfied dependencies:
    tensorflow: not installed, required: 1.2.0

 

I have installed both tensorflow and tensorflow-gpu. I also ran the install_requirements scripts, but it did not help. I tried reinstalling OpenVINO, but without any luck. I can easily import tensorflow if I open a python3 terminal.

Anyone experienced similar issues?

 

System information:

Ubuntu 16.04

Tensorflow 1.12.0

Python 3.5.2

 

Thanks,

Aske

0 Kudos

Accepted Solutions
nikos1
Valued Contributor I
161 Views

Hi Aske,

> sudo python3 mo.py --input_model /home/asrl/openvino_models/mobilenet_v1_0.5_160_frozen.pb

What  happens if your try without sudo? ( sudo can sometimes/always get you in a different OpenVino environemmt. )

for example

source  ~/intel/computer_vision_sdk/bin/setupvars.sh
python3 mo_tf.py --input_model  ...

cheers

nikos

View solution in original post

2 Replies
nikos1
Valued Contributor I
162 Views

Hi Aske,

> sudo python3 mo.py --input_model /home/asrl/openvino_models/mobilenet_v1_0.5_160_frozen.pb

What  happens if your try without sudo? ( sudo can sometimes/always get you in a different OpenVino environemmt. )

for example

source  ~/intel/computer_vision_sdk/bin/setupvars.sh
python3 mo_tf.py --input_model  ...

cheers

nikos

View solution in original post

Lejboelle__Aske
Beginner
161 Views

nikos wrote:

Hi Aske,

> sudo python3 mo.py --input_model /home/asrl/openvino_models/mobilenet_v1_0.5_160_frozen.pb

What  happens if your try without sudo? ( sudo can sometimes/always get you in a different OpenVino environemmt. )

for example

source  ~/intel/computer_vision_sdk/bin/setupvars.sh
python3 mo_tf.py --input_model  ...

cheers

nikos

 

It works. I, of course, just had to also specify an output dir in my home folder since the toolkit is installed in /opt/intel/.

Thanks.