引導式測驗 · 中等 等級
反監控時尚與臉部辨識迴避測驗
區分人臉檢測和識別,並了解為什麼對抗性化妝或服裝僅在特定測試條件下有效。
問題 1 的 8
原始 CV Dazzle 專案針對什麼?
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。
- 原始 CV Dazzle 專案針對什麼?
- The Viola–Jones face detector
- Every current facial-recognition model
- GPS追蹤
- A fingerprint scanner
- 為什麼偽裝圖案在另一套相機系統上可能失效?
- All detectors use the same code
- The technique is system- and environment-specific
- Patterns never affect images
- Cameras do not capture faces
- 臉部偵測與臉部辨識有何不同?
- Detection identifies a person; recognition only detects motion
- They are identical stages
- Detection locates a face; recognition compares it to identities or stored images
- Recognition happens before a face is captured
- 2016 年的對抗眼鏡研究證明了什麼?
- All public cameras can be defeated
- Glasses erase location history
- Infrared cameras cannot see faces
- Patterned glasses affected specific recognition models in experiments
- CV Dazzle 能保證免於現代監控嗎?
- Yes, it blocks every sensor
- No; its original looks targeted an older detector and current effects vary
- Only when worn indoors
- Yes, if makeup is bright
- Why distinguish false rejection from misidentification?
- They are the same metric
- False rejection always causes arrest
- Mistaking someone for another person can create additional harms
- Misidentification cannot happen
- Which variables should a real-world evaluation vary?
- Only the makeup brand
- Only the wearer’s name
- The number of social media followers
- Camera, distance, lighting, angle, motion, and model version
- Which privacy measure is more reliable than a single adversarial clothing pattern?
- Limit collection, retention, access, and deployment of facial data
- Assume no camera can see the wearer
- Use one makeup design in all lighting
- Rely only on infrared-blocking glasses