Help & Contact

Yes. The 20-20-20 timer, full-screen break overlay, and break-completion stats all keep working without the camera. Blink rate, PERCLOS, posture, and distance metrics will simply be unavailable. To enable them later: System Settings → Privacy & Security → Camera → toggle EyeGuard on, then restart the app.

是的。无摄像头时,定时提醒、休息遮罩、休息达成统计都正常工作。如要启用监测,在系统设置打开摄像头权限后重启 app。

Validated against Google MediaPipe Face Landmarker (an industry research benchmark), EyeGuard's EAR-based detection achieves F1 = 78.6% (precision 78.6%, recall 78.6%) under typical desk lighting. The Vision 6-point eye contour caps the algorithm around 80–85% F1; we're shipping a CoreML eye-state classifier in v1.5 to push past 90%.

For a single sample window, treat the rate as directional (am I blinking too little?) rather than absolute. Trends over a week are far more meaningful than any single number.

用 MediaPipe 校准后 F1=78.6%。Vision 6 点 contour 是天花板。v1.5 加 CoreML 模型可破 90%。建议看趋势而非单次绝对值。

Preferences → Stats → Clear all data. A confirmation dialog appears; clicking through wipes the entire SQLite database immediately. Or quit EyeGuard and remove the entire sandbox folder:

rm -rf ~/Library/Containers/com.alex.eyeguard

Either method removes 100% of EyeGuard data. There is no remote copy, by design.

偏好 → Stats → 清除所有数据;或终端运行 rm -rf ~/Library/Containers/com.alex.eyeguard。无云端副本。

Distance estimation uses a personal calibration constant (K) derived from your face-box width at a known sitting distance. If you skipped or rushed the calibration wizard, the default K assumes a 720p FaceTime camera at ~40cm — that can be off for laptops with non-standard cameras or external displays.

Re-run calibration: Preferences → Monitoring → Recalibrate distance…. The 3-step wizard takes about 15 seconds and noticeably improves accuracy.

距离不准是因为没校准。偏好 → 监控 → 重新校准距离,3 步 15 秒搞定。

EyeGuard is sold through the Mac App Store. Refunds are handled by Apple under their Standard Terms — request one through reportaproblem.apple.com within 14 days of purchase. We don't see your purchase information directly, so we can't process refunds on our side, but we'll happily help you frame the request.

通过 Mac App Store 购买,退款走 Apple reportaproblem.apple.com,14 天内可申请。我们看不到你的购买信息,但可以协助你写申请。

Three common causes:

• Lighting. Strong backlight (window behind you) makes Vision struggle to detect facial landmarks. Sit so the brightest light source is in front of you.
• Glasses with strong reflections. Anti-glare coatings help; tilting your head slightly forward can also reduce reflections.
• Camera angle. Vision performs best on faces within ~20° of straight-on. If your camera is mounted very high or low, blink detection accuracy drops.

If problems persist, run a manual check (menu bar → Test now) and check the result card's "Face coverage" metric. Anything below 70% indicates the camera isn't seeing you reliably.

主要原因: 逆光、强反光眼镜、摄像头角度过偏。手动检测看 "人脸覆盖率",低于 70% 就是摄像头没看清你。

No. EyeGuard is an "accessory" app (LSUIElement = YES) and lives only in the menu bar by design. To quit, click the menu bar icon → Quit EyeGuard, or press ⌘Q while the popover is open.

不会。EyeGuard 是菜单栏 app,Dock 不显示。退出: 菜单栏图标 → 退出,或 popover 内 ⌘Q。

EyeGuard is a universal binary (x86_64 + arm64), so it runs on both Intel and Apple Silicon Macs. We test on Apple Silicon — performance is noticeably better there because Vision uses the Neural Engine. On Intel, expect higher CPU usage during sampling windows (still negligible at idle, since the camera is off most of the time).

通用二进制,Intel + Apple Silicon 都能跑。Apple Silicon 上 Vision 走神经引擎,性能更好。