GCP Cloud Digital Leader Practice Question

Your team must add image classification to a new mobile app. The business needs a custom model that recognizes novel product categories. Engineers have ML expertise and several weeks for data prep, training, and tuning. Which Google Cloud approach best fits these requirements and trade-offs?

Upload a small sample of images to AutoML Vision and rely on its automatic model tuning to meet the requirements with minimal effort.
Run a BigQuery ML built-in model to classify images stored in Cloud Storage using a simple SQL query.
Use Cloud Vision pre-trained API to detect labels in uploaded images with no additional training.
Use Vertex AI to build and train a completely custom image-classification model from your own training code and data.

Report Issue

Answer Description

Building and training a custom model on Vertex AI offers the greatest flexibility and opportunity for differentiation because you can design the model architecture, choose the training framework, and fine-tune hyperparameters using your own dataset. This requires the most ML expertise and development effort and has a longer time-to-value, but it is appropriate when you need to recognize categories that pre-trained models and AutoML cannot handle out of the box. Pre-trained Vision API is fast and requires almost no ML skill but offers limited customization. AutoML Vision provides more flexibility than a pre-trained API while still abstracting infrastructure, yet it does not grant full control over the modeling pipeline like custom training on Vertex AI. BigQuery ML does not natively support training image-classification models.

Ask Bash

Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.