Intel® Distribution of OpenVINO™ Toolkit
Community support and discussions about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all things computer vision-related on Intel® platforms.

Model Optimizer TensorFlow conversion

Stancu__Robert
Beginner
144 Views

Hi guys,

I'm having trouble converting a frozen tf model to IR. It's a Multi Column CNN arhitecture. Frozen model can be found here: https://github.com/RobertStancu/CrowdCount. Here is the repository from where I've trained and frozen the model: https://github.com/uestcchicken/crowd-counting-MCNN. 

Here's the error: shape [-1,-1,-1,1]. is not fully defined for output 0 "PlaceHolder". Same happens when trying to change the input shape with the --input_shape parameter. 

Thanks in advance.

0 Kudos
1 Reply
Shubha_R_Intel
Employee
144 Views

Dear Stancu, Robert

Though numpy is "A-OK" with -1 as values for layer shapes OpenVino is not.  OpenVino expects that layer shapes and dimensions within a model are explicitly defined and also positive (greater than 0).

Hope it helps,

Thanks,

Shubha

 

Reply