Camera wearer voice activity detection on EasyCom (full)
91.69Wearer Audio activity mAPEgocentric Deep Multi-Channel Audio-Visual Network
Evaluation Results
| Method | Links | |
|---|---|---|
| Egocentric Deep Multi-Channel Audio-Visual NetworkAudio Feature Type=Spectrogram (spec)2022.01 | 91.69 | |
| Egocentric Deep Multi-Channel Audio-Visual NetworkAudio Feature Type=Correlation (cor)2022.01 | 90.2 | |
| Egocentric Deep Multi-Channel Audio-Visual NetworkAudio Feature Type=Spectrogram (spec), Channels=22022.01 | 90.14 | |
| Egocentric Deep Multi-Channel Audio-Visual NetworkAudio Feature Type=Correlation + Energy (cor+eng)2022.01 | 90.13 | |
| Egocentric Deep Multi-Channel Audio-Visual NetworkAudio Feature Type=Energy (eng)2022.01 | 88.89 | |
| Egocentric Deep Multi-Channel Audio-Visual NetworkAudio Feature Type=Correlation (cor), Channels=22022.01 | 87.66 | |
| AV-rawaudioInput Modality=Audio-Visual2022.01 | 87.29 | |
| Eng(single channel)Feature=short-time energy, Channels=12022.01 | 76.71 |