Table. 2.
Examples of information that can be quantified from drawings
| Image-level metrics | Detail-level metrics | Computational metrics |
|---|---|---|
| Image/scene identity | Object identities | Saliency |
| Drawing quality | Feature size/location/orientation | Color/luminance |
| Aesthetic quality | Similarity to a schema/exemplar | Symmetry |
| Emotional valence | Texture and material of objects | Spatial frequency, edges, GIST |
| Viewing perspective | Caricaturization of features | Mid-level measures: SIFT, HOG |
| Realism/imaginativeness | Usage of color | Deep learning: classification & heatmaps |
| Similarity to a schema | Usage of text labeling | Meaning maps |
| Accuracy | Drawing order & time | Motion maps |
| Erasures and errors | Object-object relationships | Semantic segmentation |
| Interpretability | Insertions of false details | Generative Adversarial Network inputs |