引導式測驗 · 中等 等級
AI 增強照片與虛構細節的測驗
Test underdetermined enhancement, plausible versus observed detail, original preservation, forensic limits, disclosure, and task-specific checks.
問題 1 的 8
Why can AI enlargement invent a plausible facial feature?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- Why can AI enlargement invent a plausible facial feature?
- Many high-resolution details fit the same low-resolution pixels
- The camera stored a hidden high-resolution face
- All model outputs are historical records
- Interpolation always knows missing pixels
- What should remain available after enhancing a photo?
- Only the most attractive render
- Only a cropped thumbnail
- The unmodified original and processing record
- 僅型號提示
- An enhanced license plate appears to show a letter absent from the original. What is the safest treatment?
- Use it as certain camera evidence
- Treat the letter as a model prediction needing independent evidence
- Erase the original file
- Assume every enhancement agrees
- Why compare an enhanced image with the original at the same crop?
- To prove all new pixels are true
- To increase the model score
- To hide compression
- To see what changed in the details relevant to the claim
- What does a convincing sharpness score fail to prove?
- That a specific word or identity was reconstructed correctly
- That the output has pixels
- That some texture looks sharp
- That the file can be displayed
- When can enhancement be acceptable as an illustration?
- Whenever it is used as unmodified evidence
- Only if no source survives
- When its interpretive role is clear and material changes are disclosed
- Only if the model gives one output
- Why do multiple plausible outputs from one input matter?
- They prove all variants were photographed
- They show uncertainty about the unseen detail
- They identify the original color
- They remove the need for a source
- What should back a consequential identity or text claim?
- Only a larger export
- Only a model confidence adjective
- Only a social-media repost
- Independent source evidence beyond the enhanced single image