IBM Maximo Visual Inspection — Orchestrated Deep Learning Services
IBM Maximo visual inspection is an offering from Maximo application suite which has the power of asset management, remote monitoring and predictive maintenance. MVI utilise deep learning models imbibed computer vision capabilities for automated inspection. People with no coding background can use the tools in this offering to label data, train and deploy deep learning vision models.
In this tutorial, we will see functionalities offered by MVI. I will label data for mask detection, train, deploy and utilise the deployed model in a dashboard for live monitoring. For this tutorial I have taken my 2 minutes video with mask on and off and uploaded on MVI. MVI has capabilities of auto labelling the data set and divide video into images for training the model.
- Auto label dataset
You can simply upload a video with mask on and off and click on upload datasets. MVI provides features like customised frame capturing from video after any defined interval of time. For auto label you need a small model trained on same dataset. You can label 5–10 images manually, train model on this and use this model to auto label the entire dataset. You can augment the dataset also.
2. Train model
You get variety of options in MVI like image classification, object detection and action detection. In this blog we will use Faster — RCNN for mask detection. You can customise the hyper parameters as well.
3. Deploy model
Once the model is trained you can deploy model on MVI. Deployed model will have a API endpoint and key which will be exposed to dashboard for live monitoring and enjoy live monitoring.
4. Live monitoring
You can refer to Github code for setting up dashboard — https://github.com/IBM/maximo-visual-alerts. Pass the API key and endpoint obtained from MVI.
For more detailed setup on MVI and dashboard refer to below links —
