Google Professional Machine Learning Engineer Question 233
Single answerGoogle Cloud PlatformYou are building a machine learning model that requires processing millions of images for training. You want to store and organize these images efficiently on Google Cloud to enable scalable access during training. Which approach should you take?
- A
Store the images in a Google Cloud Storage bucket with a structured naming convention to reflect their labels.
- B
Use BigQuery to directly store and query the image files as binary data.
- C
Store the images in Firestore with metadata for each image document to describe its label.
- D
Use Google Cloud Storage buckets to store the images and BigQuery to store metadata such as labels and file paths.
Show answer and explanation
Correct answer: D
Explanation
The optimal solution for storing and organizing large datasets like images is to use Google Cloud Storage for the actual files and BigQuery for metadata management. Google Cloud Storage provides scalability and cost efficiency for object storage, while BigQuery allows you to perform fast queries on metadata, enabling better organization and access patterns during training.
- A. Incorrect.
While Google Cloud Storage is suitable for storing images, using a structured naming convention alone can make querying or managing metadata difficult when dealing with millions of files.
- B. Incorrect.
BigQuery is optimized for analytical queries, not for storing and accessing large binary files like images. This approach is not scalable for image data storage.
- C. Incorrect.
Firestore is designed for high-availability document storage, but it is not optimized for storing large binary files like images. This approach could lead to inefficiencies.
- D. Correct.
This is the best option. Google Cloud Storage is optimized for storing large binary files like images, while BigQuery can efficiently store and query associated metadata like labels and file paths for scalable access.