Table 6
Performance of networks using difference images.
| CNN | ResNet | |||||
|---|---|---|---|---|---|---|
| Variations | AUROC | TPR0 | TPR10 | AUROC | TPR0 | TPR10 |
| Baseline gri | 0.9557 | 0.0 | 44.4 | 0.9913 | 36.0 | 55.0 |
| DI only | 0.8780 | 0.5 | 4.2 | 0.9322 | 1.1 | 11.6 |
| DI + i-band | 0.9495 | 7.9 | 34.4 | 0.9530 | 3.7 | 14.3 |
| DI + gri | 0.9684 | 9.5 | 45.0 | 0.9883 | 24.3 | 42.9 |
| gri + masks | 0.9949 | 0.0 | 50.0 | 0.9559 | 7.45 | 14.9 |
Notes. The middle three rows list networks trained on difference images from the baseline data set, with metrics evaluated on the difference images from the test set. The last networks use the original gri stacks, after masking surrounding galaxies within the cutouts. Further details are given in Sections 5.7 and 5.8.
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.