Open Access

Table C.1

Mean average precision (mAP) calculated at different intersection-over-union (IoU) thresholds for base architecture networks trained from scratch on either RGZ or MGCLS.

IoU RGZ mAP MGCLS mAP
ViT-Base supervised
0.1 0.522 0.901
0.2 0.519 0.901
0.3 0.517 0.900
0.4 0.514 0.0899
0.5 0.510 0.885
0.6 0.492 0.792
0.7 0.438 0.437
0.8 0.318 0.095
0.9 0.079 0.018
ResNet 50 supervised

0.1 0.662 0.888
0.2 0.661 0.888
0.3 0.661 0.888
0.4 0.659 0.887
0.5 0.654 0.874
0.6 0.645 0.808
0.7 0.598 0.469
0.8 0.480 0.096
0.9 0.142 0.012
ResNet 18 supervised

0.1 0.658 0.720
0.2 0.657 0.696
0.3 0.655 0.657
0.4 0.651 0.601
0.5 0.641 0.511
0.6 0.614 0.343
0.7 0.555 0.140
0.8 0.415 0.021
0.9 0.108 0.011

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.