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

Learning with intel caffe.

Ha__Seongwook
Beginner
391 Views

Hi.

I used Intel's license-plate-recognition, person-vehicle-bike detection, vehicle-attributes-recognition, and vehicle-license-plate-detecction models. However, there is no local model.

I want to create a local image database and study with Intel Caffe.

I am going to learn with Intel Caffe based on GPU-less Xeon Processor.

Where can I get a prototxt file for learning and testing of the supplied model to use the intel inferencing engine's capabilities as a whole?

Best regards.

Seongwook, Ha.

0 Kudos
4 Replies
Monique_J_Intel
Employee
391 Views

Hi Seongwook,

The pre-trained model's "trained" prototxt files are not available currently but there is future discussion on possibly exposing these. Currently, the "deployment" prototxt files are included in the latest OpenVINO package for the pre-trained models. Let me know if you have any further questions.

Kind Regards,

Monique Jones

0 Kudos
Monique_J_Intel
Employee
391 Views

Hi Seongwook,

Another option that you have is to replace the input layer in the prototxt files provided in the OpenVINO package for a particular model to accept the data from your training dataset, add/replace output layer to layers with a loss function, and define local learning rates and introduce weight initialization.

Kind Regards,

Monique Jones

0 Kudos
Ha__Seongwook
Beginner
391 Views

Hi Monique.

Of course, I can do a variation of deployment prototxt, but train and test prototxt will be a bit different from the intel optimized model.
I want to achieve performance equal to or better than the optimization performance offered by the intel samples.
So I want to be helped by intel VNO team.

Kind Regards.

Seongwook, Ha.

0 Kudos
Seunghyuk_P_Intel
391 Views

 

Hi Seongwook,

I had a discussion with 이인구 전무님 in Intel Korea regarding this.

I will check with him and may send email to you again if he wants.

Regards,

Peter.

 

0 Kudos
Reply