Table 1.
Single class agreement results on the first minute of all seven eye-tracking videos between manual annotators and the computer vision algorithm
| Video | Duration (s) | Face-gaze duration (s) | Δ (s) | normalized Δ (%) | Cohen’s kappa | |
|---|---|---|---|---|---|---|
| Manual | Algorithm | |||||
| #1 | 59.84 | 30.44 | 30.60 | – 0.16 | 0.27 | 0.97 |
| #2 | 60.50 | 27.00 | 26.60 | 0.40 | 0.66 | 0.89 |
| #3 | 60.04 | 30.08 | 31.40 | – 1.32 | 0.22 | 0.94 |
| #4 | 60.46 | 35.36 | 36.56 | – 1.20 | 1.98 | 0.96 |
| #5 | 60.20 | 24.80 | 26.00 | – 1.20 | 1.99 | 0.98 |
| #6 | 60.00 | 24.64 | 26.68 | – 2.04 | 3.40 | 0.91 |
| #7 | 60.00 | 34.4 | 34.56 | – 0.16 | 0.27 | 0.90 |