Google Exams and Introduction to Vertex AI and AutoML Tabular

In addition to the AutoML capabilities of the cloud platform, GCP also offers pre-trained models. Pre-trained models belonging to Natural language, vision, videos, translation, and so on, are available. Predictions from the pre-trained models can be obtained from the readily available APIs (without users deploying the models).

Under the Artificial Intelligence section of the navigation menu, these capabilities are listed as shown in Figure 3.50:

Figure 3.50: Artificial Intelligence section in the navigation menu of GCP

1. Pre-built models for Natural Language.
2. Pre-built models for images.
3. Pre-built models for videos.
4. Pre-built models for translation.

Clicking on the Natural Language will navigate to the page as shown in Figure 3.51:

Figure 3.51: Natural language landing page

1. Natural language pre-trained models:
   1. They can be used for various tasks such as sentiment analysis, entity analysis, entity sentiment analysis, content classification, and syntax analysis.
2. Healthcare Natural Language APIs can be used for deriving insights from medical text.

Clicking on Vision will navigate to the page as shown in Figure 3.52:

Figure 3.52: Vision landing page

1. Vision pre-trained models:
   1. They can be used for mage labelling, face and landmark detection, optical character recognition, and so on.
2. Retailers can design products with reference photos that individually visually explain the product from a variety of angles using Vision API Product Search.

View API docs contain sample code in Python, Java, Go, and other languages which can be directly used to get the predictions from these pre-trained models. The sample code does not contain a piece of code for user authentication. Users can add the same line of code that has been used for authentication while getting the predictions from the AutoML models.

Note:

• Users need to install google-cloud-language for natural language models and google-cloud-vision for vision models.
• Pre-trained models are trained with generic data and may not give accurate predictions for business-specific problems.
• Figures shown in the pre-trained models also contain the AutoML section, that is because before Vertex AI AutoML was a separate section in GCP.