Child Drowsiness and Fatigue Detection

Problem Solved

In classrooms and home study setups, signs that a child is drowsy during homework or online lessons can be missed. Raw video review is subjective and affected by posture, lighting, and camera angle. This skill focuses on child learning scenarios and turns facial-video cues—eye closure, head nods, and eye-region glossiness changes—into trackable fatigue indicators.

How It Works

Inputs can be local mp4/avi/mov files or network URLs, with a frontal face, at least 15 FPS, and no more than 10MB. Core capabilities include:
- Detecting child face state and classifying eyes as open / half-closed / closed
- Computing PERCLOS, continuous eye-closure duration, and blink frequency
- Counting head nods and angles, and observing changes in eye-region glossiness
- Outputting a 0-100 fatigue index and alert / mild_fatigue / moderate_fatigue / drowsy levels
- Producing drowsiness event lists, risk notes, rest reminders, and report links

For historical report lookup, the skill must read from the cloud API rather than local memory or manually compiled summaries, and results are usually rendered as a Markdown table.

Boundaries

This is best suited for fixed-camera settings such as classrooms, home desks, and online classes. It is an auxiliary learning-fatigue signal, not a replacement for parental or teacher observation, and it should not be used as medical diagnosis or sleep-disorder diagnosis. If a child shows persistent severe drowsiness, consult a qualified clinician. Because the input involves minors, guardian consent and secure storage of video data are required.