Table 2.
Correlation of top-four models in each of the three superordinate-level scene categories. The function-based model performs similarly in all types of scenes, while the CNN, attribute, and object-based models perform poorly in indoor environments.
Indoor | Urban | Natural | |
---|---|---|---|
Functions | 0.50 | 0.47 | 0.51 |
CNN | 0.37 | 0.43 | 0.59 |
Attributes | 0.15 | 0.20 | 0.41 |
Objects | 0.19 | 0.27 | 0.44 |