Table 8 Results (%) from several multi-task models trained to predict different tasks.
From: A new periocular dataset collected by mobile devices in unconstrained scenarios
Model | Rank 1 | Rank 5 | Device model | Age | Gender | Eye side |
|---|---|---|---|---|---|---|
Multi-task (no model) | \(80.76\pm 0.94\) | \(91.96\pm 0.51\) | – | \(82.14\pm 0.83\) | \(97.72\pm 0.17\) | \(\mathbf {99.99\pm 0.01}\) |
Multi-task (no age) | \(81.93\pm 0.99\) | \(93.51\pm 0.69\) | \(87.20\pm 0.63\) | – | \(97.65\pm 0.20\) | \(\mathbf {99.99\pm 0.01}\) |
Multi-task (no gender) | \(82.48\pm 0.64\) | \(93.55\pm 0.52\) | \(86.71\pm 0.54\) | \(83.17\pm 0.54\) | – | \(\mathbf {99.99\pm 0.01}\) |
Multi-task (no side) | \(83.72\pm 0.61\) | \(94.07\pm 0.54\) | \(87.22\pm 0.79\) | \(83.75\pm 0.53\) | \(97.70\pm 0.20\) | – |
Multi-task | \(\mathbf {84.32\pm 0.71}\) | \(\mathbf {94.55\pm 0.58}\) | \(\mathbf {87.42\pm 0.65}\) | \(\mathbf {84.34\pm 0.71}\) | \(\mathbf {97.80\pm 0.21}\) | \(99.98\pm 0.02\) |
