You need to detect PPE compliance (masks or safety glasses) on a periodic schedule (every 15 minutes) while minimizing development effort and cost.

Why Computer Vision (Azure AI Vision)Azure AI Vision (formerly “Computer Vision”) lets you analyze still images with prebuilt Image Analysis and—when needed—train a lightweight custom object detection model directly in Vision Studio without building ML pipelines. You can simply grab a frame (snapshot) from your camera feed every 15 minutes and call the Vision API—this keeps costs low because you pay per image analyzed rather than continuously processing video. Microsoft’s current Vision stack supports image analysis and custom object detection via Vision Studio and v4.0 APIs, designed for scenarios like PPE detection. Microsoft Azure+1

Why not FaceThe Face service is intended for face detection/recognition workflows. While Face can return some accessory attributes (e.g., glasses/mask confidence), Microsoft has scaled back several face attributes under Responsible AI guidance, and Face is not positioned as a general PPE-compliance detector across a scene (multiple people, full body PPE). It’s therefore a poorer fit and can increase false decisions for PPE checks beyond the face region. Azure Docs+3

Why not Azure Video Analyzer for Media (Video Indexer)Video Indexer is optimized for extracting insights from full videos (speech, people, labels, brands). Running it just to check PPE every 15 minutes is overkill and more expensive than analyzing a single still frame with Vision. In addition, Microsoft has been adjusting video services post–Azure Media Services retirement; using snapshot + Vision avoids those complexities and keeps costs minimal.

How this meets the requirements

Identify masks/safety glasses: Use Image Analysis or a quick custom object detector in Vision Studio trained on a small set of PPE images. Microsoft Learn

Every 15 minutes: Capture a single frame from the camera and call the Vision endpoint—simple scheduling with a timer-triggered Function or Logic App, minimal code. (General Azure pattern; Vision API is stateless per image.) Microsoft Azure

Minimize development effort: Vision Studio provides no-code/low-code training and testing for custom detectors if the prebuilt model isn’t sufficient. Microsoft Learn

Minimize costs: Pay-per-call image analysis is cheaper than indexing entire video segments; no continuous streaming required. Microsoft Azure

Answer: B. Computer Vision

Microsoft References

Azure AI Vision (product overview): image and video analysis with OCR, object detection, Image Analysis. Microsoft Azure

What’s new in Azure AI Vision: custom image classification/object detection in Vision Studio and v4.0 APIs. Microsoft Learn

Face detection & attributes (accessories including glasses and mask)—not the right tool for scene-level PPE compliance. Microsoft Learn

Face Detect REST API (attributes list, including glasses/mask). Microsoft Learn

Responsible AI updates indicating changes to facial analysis attributes; facial detection/accessories remain but are limited in scope. Azure Docs

Azure AI Video Indexer release notes/updates (video-centric use cases, ongoing adjustments post–AMS retirement).