Table 8 Statistics of monocular TcNet01 on two Rubin Face-vase-like rectangle-window datasets. The mean average precision (mAP) calculations follow the criteria outlined in sect. “Statistical evaluation criteria”. (a) monocular TcNet01 trained on the ‘Darker-region-as-object’ dataset; (b) monocular TcNet01 trained on the ‘Lighter-region-as-object’ dataset.
From: Understanding and simulating border ownership centered segmentation
TcNet01 | Darker-region-as-object (inference) | Lighter-region-as-object (inference) | ||
|---|---|---|---|---|
All contour | Border ownership | All contour | Border ownership | |
(a) | 97.8% | 95.1% | 94.0% | 0% |
(b) | 88.4% | 0% | 99.4% | 98.5% |
