This project provides an automated office activity tracking system using YOLOv8 for detection and DeepSORT for multi-object tracking.
Each employee is detected, tracked across frames, and classified into one of six activity states.
The system calculates working vs. non-working time, logs it in a CSV file, and saves annotated videos.
⬤ Real-time activity detection using YOLOv8
⬤ Consistent employee tracking with DeepSORT
⬤ Calculates total working and idle time for each employee
⬤ Saves activity summary as CSV logs
⬤ Exports annotated output video with live tracking
⬤ Color-coded bounding boxes:
- 🟢 Green → Working
- 🔴 Red → All other classes (Not Available, Not Working, Roaming, Sleeping, Talking.)
✅
| Class ID | Activity Label | Description |
|---|---|---|
| 0 | Not Available | No valid detection |
| 1 | Not Working | Idle or not engaged |
| 2 | Roaming | Moving without active work |
| 3 | Sleeping | Inactive or resting |
| 4 | Talking | Engaged in conversation |
| 5 | Working | Actively performing tasks |