Data Preparation for Model Training

Correct Answers: A and D

AI Builder provides prebuilt and custom models in three categories: Text, Prediction, and Vision.

You can train the following model types: Text category: Category classification and Entity extraction.

Prediction category: Prediction.

Vision category: Form Processing, Object Detection.

Three of these types don't have prebuilt models: Prediction, Form Processing, Object Detection.

Before you train a model, you need to prepare the data.

AI Builder is using the Dataverse database to store the training and model data.

Some model types have their unique requirements for creating a model, like Form Processing model needs at least 5 form documents for training and Object Detection at least 15 pictures per object.

Option A is correct because you need to create a custom Dataverse table to store data for a Prediction model before training the model.

You can import data into a custom table in Dataverse using the correct format.

Option D is correct because you can have not more than two times the images for one object in the training set to train the two objects detection model.

You always need to make sure that your data is balanced.

Option B is incorrect because Microsoft recommends at least 5 form documents to train the Form Processing model.

Option C is incorrect because the training data must be in a Dataverse table in the correct format before training the custom Category classification model.

For more information about data preparation for custom AI builder models, please visit the below URLs:

Sure, I'd be happy to help!

To create a custom AI Builder model, there are several steps involved, including data preparation for the model training. The correct statements about the data preparation for the different types of models are:

A. You need to create a custom Dataverse table to store data for a Prediction model

This statement is correct for a Prediction model. To create a custom Prediction model, you will need to store data in a custom Dataverse table that contains the data you want to use for training the model. You can then use this table as a data source for the model. Once you have created the table, you can then import the data into AI Builder to train the model.

B. You need at least 10 form documents to train the Form Processing model

This statement is correct for a Form Processing model. To create a custom Form Processing model, you will need to upload at least 10 form documents that contain the fields you want to extract. The more documents you upload, the more accurate the model will be. Once you have uploaded the documents, you can then train the model in AI Builder.

C. For a custom Category classification model, you can submit data during training

This statement is correct for a Category classification model. To create a custom Category classification model, you can submit data during the training process. You will need to provide examples of the categories you want the model to classify, along with examples of data that belong to each category. You can then train the model in AI Builder using this data.

D. It is OK if you submit two times more images of one object to train the two objects Detection model.

This statement is not correct for a Two Objects Detection model. To create a custom Two Objects Detection model, you will need to upload images of the two objects you want the model to detect. You should provide an equal number of images for each object to train the model accurately. Submitting two times more images of one object can cause the model to be biased towards that object, resulting in inaccurate detection results.

I hope this explanation helps you understand the correct statements about data preparation for the different types of AI Builder models.