Explanation:

According to the Microsoft Azure AI Fundamentals (AI-900) official study guide and the Microsoft Learn module "Describe core concepts of machine learning on Azure", labeling is the process of assigning correct output values (labels) to training data before model training. In supervised learning, every input in the dataset must be paired with its corresponding output so the algorithm can learn the relationship between the two.
In this scenario, the task is to assign classes to images before training a classification model-for example, marking images as "cat," "dog," or "bird." This process defines the target variable (label) that the model will later predict. During training, the classification model uses these labeled examples to learn patterns and distinguish between categories.
Microsoft's official materials clearly define labeling as:
"The process of tagging data with the correct answer so that the model can learn to make predictions." Labeling is a crucial early step in the machine learning lifecycle, especially for image classification and natural language processing (NLP) tasks. Without accurate labels, the model cannot learn correctly and its predictions will be unreliable.
Let's briefly clarify why the other options are incorrect:
* Evaluation refers to testing the model after training to measure accuracy or performance using metrics like precision, recall, or F1 score.
* Feature engineering involves creating or selecting the most relevant input features from raw data but does not involve tagging output labels.
* Hyperparameter tuning adjusts parameters (like learning rate or depth of a tree) to optimize model performance after labeling and training have begun.
Thus, assigning classes to images prior to model training is definitively a Labeling task.