Running inference session

In this tutorial, you will learn how to run an inference session with mpVision.

After successfully training your AI model and calibrating it, you can now run an inference session to generate reports that will give you insights about your image data.

Default inference outputs of mpVision will produce a report that contains morphological statistics of the objects detected in the images. This includes the number of objects detected and their corresponding classification, area, perimeter, equivalent circular diameter, and other parameters.

Creating a dataset

After logging in to the application, click on the "Datasets" under the "Inference" navigation item in the sidebar menu located on the left side of the application view. It will navigate you to the view containing the list of your inference datasets (will be empty if you haven't created any yet).

Similar to the training datasets, you can create a new inference dataset by clicking on the "Create/import dataset" button located in the top left corner of the lists view.

For consistency, we will use the same Kaggle dataset that we used in training tutorial. That way, we can see the results of the inference session on the same data we used for training and calibration.

Initiating inference session

Upload two images from the dataset and select "Start inference" from the top left corner of images gallery view. You will see a pop-up dialog with a select input field where you can choose the model you want to use for inference. Select the model you trained in the previous tutorial and click "Start inference" button.

Similar to training and calibration sessions, the inference session speed for this tutorial will be dependent on your internet connection, and the status can be tracked in the "Reports" view under the "Inference" navigation item or in the "Sessions" popup dialog that will become available on the right side of the "Start inference" button.

When the inference session is complete, you can view the generated report by clicking on the report title in the list of reports. This will navigate you to the view containing the report details and the gallery of images with detected objects.

As displayed in the example video above, the top of the report view has a table that contains the number of AI detections per label along with the average values of statistics for each label. Below the table, you can see histograms of the statistics and the gallery of images with detected objects. For more detailed information about the report data outputs, please refer to the corresponding section in the User manual.

And that's it! You have successfully run an inference session with mpVision and generated a report. You can click on the "Download report" button to download the report in PDF format. Reports can be shared with colleagues, leveraged during audits, or incorporated in research and manufacturing initiatives.

So far we have been manually uploading images, calibrating, and running inference sessions. This is the initial user workflow to first train your AI models. But the power of mpVision comes with its automation pipelines that can be set up to run inference automatically. This automation is accomplished by connecting your devices with the application's API. You will find out how to automate mpVision in the next tutorial.