How to retrain models present in Intel's Pre-trained Models

[naman]

How models like semantic-segmentation-adas-0001 and SSD Mobilenet Caffe model(object detection) as showcased in ncapzoo and available as pre-trained models in OpenVINO can be retrained on a custom dataset? Do training extensions of OpenVINO support these models? I need to deploy the retrained models on Neural Compute Stick.

[Vladimir_Dudnik]

@naman you may check which models supported by OpenVINO Training Extensions on github

Models

After installation, you are ready to train your own models, evaluate and use them for prediction.

• Action Recognition
  • Custom Action Recognition
  • Gesture Recognition
• Instance Segmentation
  • COCO instance segmentation
  • Custom instance segmentation
• Image classification
  • Custom image classification
• Object Detection
  • Custom Object Detection
  • Face Detection
  • Horizontal Text Detection
  • Person Detection
  • Person Vehicle Bike Detection
  • Vehicle Detection
• Text Spotting
  • Alphanumeric Text Spotting

[naman]

Thanks @Vladimir_Dudnik for the reply. Do the models like efficientnet_b2b-mask_rcnn-576x576 or efficientnet_b2b-mask_rcnn-480x480 are supported on Neural Compute Stick 2? This link suggests it will not. Also in the device support, link  efficientnet_b2b is not mentioned.

I want to train and deploy a semantic segmentation model on Neural Compute Stick 2. Do you have any suggestions? 

[Vladimir_Dudnik]

@naman models efficientnet_b2b-mask_rcnn-576x576 or efficientnet_b2b-mask_rcnn-480x480 not included into Open Model Zoo, that's why they not listed in device support list. From public semantic segmentation models, deeplabv3 model included into Open Model Zoo and tested on Myriad X device.

[Syamimi_Intel]

Hi naman mehta,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question. 

Regards,

Syamimi