Input models for model optimizer

[m_v__vyshak]

Hi

I am using model optimizer to produce a valid Intermediate Representation (.xml and .bin). I want to generate .xml and .bin for the sample “interactive face detection” available within the open vino toolkit

Eventhough there are pre trained models available (face-detection-adas-0001), how can i generate the .xml and .bin files using the command

python3 mo.py –input_model <path_to_model>.caffemodel –input_proto ~/opt/intel/computer_vision_sdk_2018.3.343/deployment_tools/intel_models/face-detection-adas-0001/face-detection-adas-0001.prototxt

I could find the .prototxt file. The question is where can i get the input model files (.caffemodel or .tensorflow). Since i am new to this, it would be helpful if i could get the path or way to get these input model files

Additionally i tried the command

python3 mo.py –input_model /opt/intel/computer_vision_sdk_2018.3.343/deployment_tools/intel_models/face-detection-adas-0001.caffemodel –input_proto /opt/intel/computer_vision_sdk_2018.3.343/deployment_tools/intel_models/face-detection-adas-0001/face-detection-adas-0001.prototxt

This command threw me the following error :

[ ERROR ] The "../intel_models/face-detection-adas-0001.caffemodel" is not existing file

[Sreedhar_V_Intel]

Hi Vyshak,

If you install latest OpenVino tool kit, all pre trained models will be copied to /home/youruserid/openvino_models

[mahinlma]

This directory /home/openvino_models has only Squeezenet caffe model.

[Sandhu__Ranjit]

Morning,

I am also facing same problem as Natarajan,So any idea how to get those models ?

~ranjit

[Sandhu__Ranjit]

I think we need to run script under ..\deployment_tools\model_downloader\downloader.py yo get the models ...

~ranjit

[Mc__Andrew]

This doesn't seem to be explained anywhere but you can download them from here: https://download.01.org/opencv/2019/open_model_zoo/R1/models_bin/

[Shubha_R_Intel]

Thank you Sandhu, Ranjit and Mc, Andrew, for answering the question. Actually the download site for the models is mentioned in the actual sample documentation, for example Action Recognition Demo . If you use downloader.py pass in the --all switch which will download all the models. 

Thanks,

Shubha