Introduction

This article describes how to deploy a ModelPack model to the Maivin after training, and then run the model on the Maivin.

This article is part of a series of articles documenting the end-to-end workflow for dataset capture, annotation, model training, and deployment back to the Maivin hardware.  Refer to Maivin Dataset Overview for more details.

Deployment

1. Obtain a new model. This can be done multiple ways
   • After training a model, Quantize and Convert the model to RTM
   • Downloading a model from ModelPack releases on Github
     
     • https://github.com/DeepViewML/playingcards
     • https://github.com/DeepViewML/peopledetect
     • https://github.com/DeepViewML/facedetect
2. Copy the model onto the target

   scp ./model.rtm torizon@verdin-imx8mp-XXXXXXXX:~/model.rtm

3. Edit the /etc/default/detect file to change the model to the newly copied model
   Change the line that has MODEL = "xxxxxxxx" into:
   

   MODEL = "/home/torizon/model.rtm"

4. Restart the detect service:

   sudo systemctl daemon-reload
   sudo systemctl restart detect

5. Now the detect service is running using the new model!

Next Steps

The EdgeFirst Live Visualizer can be used to view the results of the model live in a web browers

A deployed model's results can be captured with the Maivin Dataset Recorder into the annotation stream. This annotation stream can be published to datasets and audited as part of the model and dataset lifecycle.

• Collecting Inference Annotations in the Field
• Maivin Live Camera and Detection UI