Human activity recognition leaderboard
Benchmark results for human activity recognition from wearable sensors.
Human activity recognition (IMU)
Dataset: capture-24 · Metric: Cohen's κ ↑
| Rank | Method | Cohen's κ |
|---|---|---|
| #1 | HARNet / OxWearables SSL | 0.86 |
Human activity recognition (IMU)
Dataset: capture-24 · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | ActiNet | 0.818 |
Human activity recognition (IMU)
Dataset: deo · Metric: Bal. accuracy ↑
| Rank | Method | Bal. accuracy |
|---|---|---|
| #1 | Diffusion-Based Synthetic Data Pretraining for E | 90.6 |
Human activity recognition (IMU)
Dataset: har-6bench-avg · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | Bio-PM | 0.65 |
Human activity recognition (IMU)
Dataset: mhealth · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | ZARA | 86.1 (%) |
Human activity recognition (IMU)
Dataset: opentslm-har-cot · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | Lapras | 73.53 (%) |
Human activity recognition (IMU)
Dataset: opportunity · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | ZARA | 92.5 (%) |
Human activity recognition (IMU)
Dataset: pamap2 · Metric: Macro-F1 ↑
Human activity recognition (IMU)
Dataset: prosi · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | Take it Personally | 0.64 |
Human activity recognition (IMU)
Dataset: realworld-har · Metric: Macro-F1 ↑
| Rank | Method | Macro-F1 |
|---|---|---|
| #1 | ActiNet | 0.661 |
Human activity recognition (IMU)
Dataset: udpad · Metric: Accuracy ↑
| Rank | Method | Accuracy |
|---|---|---|
| #1 | μDPad | 83.7 |