Table 2.
Bayes classification accuracy per dimension. Classifiers tested using LOOCV on individual images.
| Predictor dimension(s) | |||||||
|---|---|---|---|---|---|---|---|
| Predicted dimension | Chance | Prior-only | Semantic | Structure | Appearance | Both (naïve) | Both (non-naïve) |
| Semantic | 16.67% | 37.36% | – | 53.47% | 49.72% | 55.97% | 57.36% |
| Structure | 25.00% | 31.53% | 60.00% | – | 44.72% | 58.19% | 57.92% |
| Appearance | 20.00% | 35.69% | 43.75% | 42.22% | – | 45.83% | 50.56% |