Counts rope skipping jumps from an RTSP camera feed using YOLOv8 pose detection. Sends webhook notifications to Home Assistant on milestones.
- Real-time jump detection using head position tracking
- Rhythm validation to filter out walking/standing movements
- Distance-independent detection (works close or far from camera)
- Webhook notifications for milestone achievements
- Debug video stream with pose overlay
Copy .env.example to .env and set:
# RTSP stream URL from your camera
RTSP_URL=rtsp://user:pass@camera-ip:554/stream
# Home Assistant webhook URL for notifications
WEBHOOK_URL=http://homeassistant.local:8123/api/webhook/your-webhook-id
# Announce every N jumps (default: 100)
MILESTONE_INTERVAL=100
# Timezone for daily reset
TZ=Europe/Amsterdam# Production (real camera)
docker compose up -d rope-counter
# Test mode (loops a video file)
# Place your test video as camera-stream.mp4
docker compose --profile test up -dPre-built image available on Docker Hub: raregoat8804/opencv-rope-jump-counter
Open http://localhost:8080/debug in a browser to see the live video stream with pose overlay, jump counters, and reset button.
| Endpoint | Description |
|---|---|
GET /health |
Health check |
GET /status |
Full system status |
GET /count |
Current jump counts |
GET /reset |
Reset all counters |
GET /snapshot |
Capture current frame |
GET /debug |
Debug page with video stream, stats, and controls |
GET /debug/stream |
Raw MJPEG stream with pose overlay |
POST /session/start |
Manually start session |
POST /session/stop |
Manually stop session |
{
"event": "milestone",
"count": 100,
"milestone": 100,
"session_jumps": 100,
"daily_total": 250
}- YOLOv8-pose detects human and extracts keypoints
- Head position tracked using weighted average of visible keypoints (nose, eyes, ears)
- Jump amplitude normalized to person's bounding box height (distance-independent)
- Rhythm validation ensures consistent oscillation pattern before counting
- Horizontal stability check filters out walking/running