Speaker head orientation estimation on Real dataset (Session cross-evaluation)
73.2AccuracySTFT-phase model
Evaluation Results
| Method | Links | |
|---|---|---|
| STFT-phase modelTraining strategy=pre-trained on simulated + fine-tuned on real, Experimental protocol=Per Angle Classifier2026.07 | 73.2 | |
| DoV baselineExperimental protocol=Per Angle Classifier2026.07 | 65.4 | |
| STFT-phase modelTraining strategy=trained only on real data, Experimental protocol=Per Angle Classifier2026.07 | 62.6 |