Table A.2.
Fluctuations of the performances for the VGG and SC-VGG models.
| median | Q1 | Q3 | |||||
|---|---|---|---|---|---|---|---|
| [%] | VGG | SC-VGG | VGG | SC-VGG | VGG | SC-VGG | |
| AE | 88.3 | 89.3 | 86.2 | 88.6 | 88.6 | 90.3 | |
| pur | 93.3 | 93.3 | 92.5 | 92.3 | 94.8 | 94.2 | |
| GGSL | compl | 89.1 | 91.9 | 86.7 | 91.2 | 90.4 | 92.5 |
| F1 | 91.5 | 92.3 | 89.8 | 91.8 | 91.5 | 93.0 | |
| pur | 77.6 | 81.9 | 74.0 | 79.7 | 79.0 | 82.9 | |
| NGGSL | compl | 85.6 | 84.2 | 82.3 | 81.7 | 88.1 | 87.2 |
| F1 | 91.5 | 92.3 | 89.8 | 91.8 | 91.5 | 93.0 | |
| IQR | Q1 − 1.5 ⋅ IQR | Q3 + 1.5 ⋅ IQR | |||||
| [%] | VGG | SC-VGG | VGG | SC-VGG | VGG | SC-VGG | |
| AE | 2.4 | 1.7 | 85.3 | 87.8 | 89.7 | 91.4 | |
| pur | 2.3 | 1.9 | 90.7 | 90.1 | 95.6 | 95.7 | |
| GGSL | compl | 3.7 | 1.4 | 84.1 | 90.4 | 92.5 | 94.1 |
| F1 | 1.8 | 1.2 | 88.8 | 91.0 | 92.6 | 93.9 | |
| pur | 5.0 | 3.2 | 70.5 | 76.1 | 81.8 | 85.6 | |
| NGGSL | compl | 5.8 | 5.4 | 79.7 | 77.6 | 90.7 | 90.7 |
| F1 | 1.8 | 1.2 | 88.8 | 91.0 | 92.6 | 93.9 | |
Notes. Q1 and Q3 are the 25th and 75th percentiles. The inter-quartile range IQR = Q3 − Q1; the range (Q1 − 1.5 ⋅ IQR, Q3 + 1.5 ⋅ IQR) encloses the metric fluctuation within ±2.698σ. The best results are highlighted in bold. Average efficiency and GGSL estimators are graphically shown in the bottom panels of Fig. 6.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.