Human Activity Recognition on Physical heterogeneous edge devices V-C2 (test)
89.91AccuracyFedKD-NAS
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| FedKD-NASTotal Clients (K)=5, Hardware Setup=1 RockPi, 1 Jetson Nano, 3 Raspberry Pi2026.05 | 89.91 | 90.53 | 409.99 | 2.29 | 98.9 | 100 | 1.854 | 1.88 | |
| FedMDTotal Clients (K)=5, Hardware Setup=1 RockPi, 1 Jetson Nano, 3 Raspberry Pi2026.05 | 81.94 | 91.82 | 397.76 | 2.29 | 98.1 | 37.9 | 1.854 | 0.72 | |
| FedAvgTotal Clients (K)=5, Hardware Setup=1 RockPi, 1 Jetson Nano, 3 Raspberry Pi2026.05 | 80.93 | 92.64 | 392.8 | 4.25 | 97.9 | 30 | 1 | 0.31 |