What it means
Computer vision systems extract patterns from images or video to classify, detect, segment, track, or generate visual content.
Why it matters
Lighting, viewpoint, sensor quality, occlusion, and demographic representation can change results in ways laboratory tests miss.
Practical focus
- Define the approved purpose, accountable owner, users, operating environment, and acceptable risk for computer vision systems before implementation or procurement
- Implement capture standards, representative conditions, robustness testing, privacy review, and confidence-based fallback through documented architecture, policy, configuration, and change control
- Collect performance by environment and subgroup, sensor metadata, false-detection examples, and human-review outcomes and compare the results with approved requirements, risk thresholds, and expected outcomes
- Exercise this failure scenario as a planned test: the model loses accuracy under glare, low light, occlusion, or a camera configuration absent from training
Common mistakes
- Using the label computer vision systems without defining the exact system boundary, decision, data, and people in scope
- Treating a model response or vendor claim as sufficient evidence without testing the surrounding application and operating process
- Failing to preserve performance by environment and subgroup, sensor metadata, false-detection examples, and human-review outcomes in a form that lets reviewers reconstruct decisions and changes
- Testing only the normal path and never rehearsing what happens when the model loses accuracy under glare, low light, occlusion, or a camera configuration absent from training
Certification relevance
This subject appears in or supports the following certification bodies of knowledge: