safedrive-ai 0.2.0

Driver monitoring SDK - drowsiness, phone, seatbelt detection

SafeDrive AI — Driver Drowsiness Detection SDK

Real-time driver drowsiness and distraction detection.
Built on MediaPipe + MobileNetV3 (v0.1.0) with YOLOv8 multi-feature detection coming in v0.2.0.

---

Install

pip install safedrive-ai

---

Quick Start

from safedrive import DrowsinessDetector

detector = DrowsinessDetector(pipeline="mediapipe")

@detector.on_drowsy
def handle(event):
    print(f"Level {event.level}: {event.message}")

detector.run(camera=0)

---

Alert Levels

Level	State	Trigger	Action
1	WARNING	Eyes half-open 3s / yawn / head tilt	Soft beep
2	ALERT	Eyes closed 2s / head nod	Loud beep
3	CRITICAL	Face gone 2s / eyes closed 4s	Continuous alarm

Separate alerts (independent of drowsiness level):

• Phone detected → instant distraction alert
• Seatbelt removed → continuous safety alert
• Smoking detected → instant distraction alert

---

Full API

detector = DrowsinessDetector(
    pipeline          = "mediapipe",  # "yolo" coming in v0.2.0
    eye_close_seconds = 2.0,
    face_gone_seconds = 2.0,
    head_tilt_degrees = 15.0,
    detect_phone      = True,
    detect_seatbelt   = True,
    detect_yawn       = True,
    detect_smoking    = True,
    show_window       = True,
    ear_weight        = 0.5,
    cnn_weight        = 0.5,
)

@detector.on_drowsy      # DrowsyEvent
@detector.on_distraction # DistractionEvent
@detector.on_safety      # SafetyEvent
@detector.on_frame       # called every frame with annotated frame + FrameStats

detector.run(camera=0)
detector.stop()
detector.reset_alert()   # manual reset after CRITICAL

---

Roadmap

Version	Features
v0.1.0	MediaPipe + MobileNetV3 drowsiness detection ✅
v0.2.0	YOLOv8 pipeline + full multi-feature detection
v1.0.0	Ensemble mode + benchmark tool + full docs

---

Pipeline Comparison

Feature	MediaPipe (v0.1.0)	YOLO (v0.2.0)
Eyes open/closed	✅	✅
Yawn detection	❌	✅
Phone detection	❌	✅
Seatbelt	❌	✅
Smoking	❌	✅
Head pose	✅ (both use MediaPipe landmarks)	✅
Speed	~28 FPS	~24 FPS

---

License

MIT