Model Deployment to Maivin
Once validation is completed, we are ready to send the model to the target.
-
Select the DEPLOY button in the validation page to open the Export Model/Deploy Model Page.
There are two methods to deploy the model to the Maivin.
- Deploy Model using eIQ - directly deploys your trained model to Maivin using the eIQ toolkit. Must have completed a training/validation session in eIQ.
- Export the model to .rtm file and use the webapp interface to load the model. This method allows for easy switching between models on the target and using pre-trained models.
Deploying the Model using eIQ
- First, we need to add the remote device so we can directly send the model to the unit. Open the REMOVE DEVICES dialog by clicking on the button in the top toolbar.
- Click the ADD DEVICE button to open the add devices dialog. Enter a unit name and the URL of the Maivin unit. The URL for the maivin is just the host name, found on the label on the bottom of the Maivin unit.
- Once added, hover over the desired unit to be used, and click the select button.
- The green box indicates the currently selected unit. Multiple units can be added as remote devices. Verify which device you are communicating with before deploying the model. You can now close the Remote Devices window.
- Setup the parameters for the model to be sent to the target, such as Input and Output date type, quantization etc. Click the DEPLOY MODEL button to send the model to Maivin. The Model will be converted and sent to the unit.
- Open a browser window and go to the Maivin App webpage at http://verdin.imx8mp-xxxxxx to view the results of the model detection.
Deploy Model using Maivin Web Interface
- Instead of Deploying the model directly using eIQ, we can export the model instead. Setup the export settings as in the deploy to Maivin section, except click the EXPORT MODEL button to store the trained model as .rtm file.
- Connect to the camera Web App by navigating to http://verdin.imx8mp-xxxxxx where xxxxxxx is the verdin serial number located on the able on the bottom of the unit using a browser window.
- Click on the UPLOAD MODEL Icon on the top left of the webpage.
- Click to select a model, or drag the .rtm model file from the file explore to the window.
- The model will open and started up. At this point, everything has been executed and the model should be running on the target
Previous Step | Next Step |
ModelPack Validation | ModelPack Reference Manual |
Comments
0 comments
Please sign in to leave a comment.