A real estate organization processes photos of properties uploaded by their agents to an internal application. They must manually check and remove photos that contain personally identifiable informati

Pre-trained APIs such as Cloud Vision AI and Cloud Sensitive Data Protection require no model training or ML expertise and can be integrated into applications with simple API calls, making them the fastest path to implementation. For detecting PII such as passports and credit cards in images, these APIs are already trained on relevant data and are immediately available. This directly satisfies the requirement for a solution a developer can integrate as quickly as possible.

AutoML requires the organization to provide labeled training data and train a custom model, which takes significantly more time to configure and deploy than using a pre-trained API.

Custom training requires the most development effort, including data collection, model architecture design, training infrastructure setup, and ongoing maintenance.

BigQuery ML is designed for running ML models on structured data within BigQuery using SQL syntax, and is not suited for image-based PII detection in an application integration context.