license: cc-by-4.0
task_categories:
- tabular-classification
- robotics
tags:
- imu
- biomechanics
- activity-recognition
- physical-ai
- sensor-certification
- opportunity-dataset
pretty_name: OPPORTUNITY — S2S Physics Certified
size_categories:
- 1K<n<10K
OPPORTUNITY — S2S Physics Certified (v1.7.0)
Physics-certified windows from the OPPORTUNITY Activity Recognition dataset. Validated against 8 biomechanical laws using S2S.
Two certifications provided
| File | Columns | Rejection | Notes |
|---|---|---|---|
opportunity_rua_certified.csv |
Documented RUA sensor (cols 10-12) | 14.4% | More accurate |
opportunity_certified.csv |
Auto-detected columns | 12.3% | Zero-config |
Key finding: Documented sensor columns give different results than auto-detection. S2-ADL1 and S3-ADL1 show serious physics anomalies (20-22% pass rate).
Statistics (RUA documented)
| Metric | Value |
|---|---|
| Total windows | 2,371 |
| Files | 24 (4 subjects × ADL1-5 + Drill) |
| Sample rate | 30.3Hz (auto-detected) |
| Window size | 256 samples |
| Overall rejection | 14.4% |
| Worst session | S2-ADL1: 78% rejected |
| Best session | S2-ADL3: 9% rejected |
Key findings
S2-ADL1: 22% pass rate — serious sensor anomaly S3-ADL1: 20% pass rate — serious sensor anomaly S1-Drill: 79% pass — drill task cleanest S2-ADL3/4/5: 90%+ pass — cleanest sessions
These anomalies are invisible to standard statistical checks. S2S detects them through biomechanical law violations.
Columns used (RUA documented)
From official OPPORTUNITY column documentation:
- Accel: cols 10-12 (Right Upper Arm accelerometer, milli-g)
- Gyro: cols 50-52 (Right Upper Arm gyroscope)
Source dataset
Original OPPORTUNITY: https://archive.ics.uci.edu/dataset/226/opportunity+activity+recognition
4 subjects, 7 IMU sensors, 30Hz, daily activities + drill tasks.
Citation
S2S Physics Certification Engine v1.7.0 https://github.com/timbo4u1/S2S DOI: 10.5281/zenodo.18878307 pip install s2s-certify