MLS-C01 Question 97
Select 3You are tasked with building a machine learning model to classify images of animals into categories such as 'cat', 'dog', and 'bird'. Your dataset consists of high-resolution images stored in an Amazon S3 bucket. Which of the following steps should you take to extract and preprocess features for training your model using AWS services?
- A
Use Amazon Rekognition to extract labels from the images and use them as features for training
- B
Resize and normalize the images using AWS Lambda or Amazon SageMaker Processing to prepare them for training
- C
Use Amazon Comprehend to extract textual metadata from the images
- D
Leverage Amazon SageMaker Ground Truth to label the images if labels are not already available
- E
Use AWS Transcribe to convert audio from the images into text for feature extraction
Show answer and explanation
Correct answers: A, B, D
Explanation
To build a machine learning model with image data, it is necessary to preprocess the images (e.g., resizing and normalization) and extract meaningful features from them. Amazon Rekognition can help extract labels from images, and Amazon SageMaker Ground Truth can assist in labeling the images if labels are not already present. These steps ensure the dataset is prepared for training the model. Services like Amazon Comprehend and AWS Transcribe are not applicable to this scenario as they are designed for text and audio data, respectively.
- A. Correct.
Correct. Amazon Rekognition can analyze images and extract relevant labels (e.g., 'cat', 'dog') that can be used as features for training a classification model.
- B. Correct.
Correct. Preprocessing steps like resizing and normalizing images are critical for image-based machine learning models. AWS Lambda or Amazon SageMaker Processing can handle such preprocessing.
- C. Incorrect.
Incorrect. Amazon Comprehend is used for natural language processing and extracting insights from text. It is not relevant for image feature extraction.
- D. Correct.
Correct. Amazon SageMaker Ground Truth can be used to label datasets if the images are not already labeled, which is a crucial step for supervised learning.
- E. Incorrect.
Incorrect. AWS Transcribe is used to convert audio to text, which is not applicable to image datasets.