TABLE III.
Cross-modal AUROC averaged across tenfold cross-validation. Models were trained using DNSS scenes in booth and crowd data and tested on the held out sets from booth and crowd data from the same scenes. Standard error across 10 folds is indicated with ± sign.
| Test data | ||
|---|---|---|
| Train data | Booth | Crowd |
| Booth | 0.759 ± 0.027 | 0.753 ± 0.037 |
| Crowd | 0.753 ± 0.027 | 0.757 ± 0.038 |