Microsoft Certified Azure AI Fundamentals · Free Practice Question Medium
Question 37
A development team observes these patterns in their OCR implementation:
Scenario A: 90% accuracy on printed text, 75% on handwriting
Scenario B: 95% accuracy in good lighting, 60% in poor conditions
Scenario C: 85% accuracy on clean documents, 70% with background noise
Which conclusion about the Read API implementation is MOST accurate?
-
A
Environmental factors have more impact than text type
-
B
Performance variations suggest incorrect API configuration
-
C
Image preprocessing would only benefit Scenario B
-
D
The system requires separate models for each scenario
Reveal correct answer
Correct answer: A
Explanation
Impact of Environmental Factors
Lighting Conditions:
The significant drop from 95% to 60% accuracy between good and poor lighting (Scenario B) shows that lighting conditions heavily influence OCR performance. Poor lighting can lead to reduced contrast and blurred details, making it difficult for the OCR engine to correctly recognize text.Background Noise:
Similarly, in Scenario C, the decline from 85% to 70% accuracy when background noise is introduced indicates that extraneous visual information interferes with text recognition.Text Type Variations:
Although there is a noticeable difference between printed text and handwriting (Scenario A), the relative impact is less severe compared to the impact of lighting and background noise.
Conclusion Drawn from the Patterns
The overall trends strongly suggest that environmental factors—such as lighting conditions and background noise—have a greater effect on OCR accuracy than the inherent differences between printed text and handwriting. These factors degrade image quality, which is crucial for the Read API to function optimally. Thus, the most accurate conclusion is that environmental factors have more impact than text type.
Microsoft Documentation References and Further Readings
Azure Cognitive Services – Computer Vision
Microsoft's Educational Content For This Topic:
A.
Comparing the scenarios, the drop in accuracy due to lighting conditions (95% to 60%) is far more significant than the drop from printed text to handwriting (90% to 75%). This indicates that variations in environmental factors, such as lighting and background noise, exert a greater influence on the system’s performance than the type of text itself. Hence, environmental conditions are the key driver behind the performance differences observed.
B.
This option suggests that the discrepancies in accuracy are due to a misconfiguration of the API. However, the variations align with known challenges in OCR systems, where handwriting and adverse environmental conditions naturally reduce accuracy. The data does not necessarily point to an incorrect setup, but rather to inherent limitations under different conditions.
C.
While image preprocessing can significantly improve performance in poor lighting conditions (as seen in Scenario B), it can also enhance results in cases with background noise (Scenario C). This option incorrectly limits the benefits of preprocessing to only one scenario. Preprocessing generally helps improve overall image quality regardless of the specific issue.
D.
This option implies that different models are needed for handling printed text, handwriting, and various environmental conditions. However, the observed performance variations can often be addressed by tuning and enhancing a single model rather than deploying entirely separate models. The differences do not necessarily indicate that separate models are required.
Discussion
Think the marked answer is wrong, or have a better explanation? Share it below — comments appear after review.
