Microsoft Azure AI Fundamentals (AI-900)Describe features of computer vision workloads on AzureHard
A smart factory is implementing a system to detect anomalies in product packaging on an assembly line. The system needs to precisely identify areas of damage, such as tears, dents, or incorrect labeling, by highlighting the exact pixel regions affected. This requires a highly detailed understanding of the image content. Which computer vision capability is BEST suited for this level of precision?
- AObject Detection
- BImage Classification
- COptical Character Recognition (OCR)
- DSemantic Segmentation
Show answer & explanationAnswer & explanation
Correct answer: D. Semantic Segmentation
Semantic Segmentation provides pixel-level classification, allowing the system to precisely identify and highlight the exact regions (pixels) corresponding to anomalies like tears or dents on product packaging, which is crucial for detailed anomaly detection.
Why the other options are wrong
- A. Object Detection would draw a bounding box around a damaged area, but not precisely highlight the exact pixel regions of the damage.
- B. Image Classification would only tell if the overall package is 'damaged' or 'undamaged', not *where* or *what specific part* is damaged.
- C. OCR is for extracting text, not for detecting physical damage or anomalies on objects.
Semantic Segmentation for Anomaly Detection
Applying semantic segmentation to precisely outline and identify anomalous regions (e.g., defects, damage) at a pixel level within images.
- Provides highly granular localization of anomalies.
- Essential when exact shape and extent of damage are critical.
- Often used in quality control, medical imaging, and industrial inspection.
Memory trick: Segment pixels to pinpoint the precise problem.