Open Access

Table C.1

Performance vs fine-tuned layers

Backbone Name Fine-tuned Layers MB error MB F1
DINO Resnet 0 0.09 ± 0.01 0.92 ± 0.01
1 0.07 ± 0.01 0.93 ± 0.01
2 0.08 ± 0.01 0.92 ± 0.01
3 0.09 ± 0.01 0.92 ± 0.01
4 0.07 ± 0.00 0.93 ± 0.00
5 0.08 ± 0.01 0.92 ± 0.01
6 0.09 ± 0.00 0.91 ± 0.01
7 0.08 ± 0.00 0.93 ± 0.00
8 0.07 ± 0.00 0.93 ± 0.00
9 0.09 ± 0.01 0.92 ± 0.01

GZ2 MoCo 0 0.09 ± 0.00 0.91 ± 0.00
1 0.08 ± 0.00 0.93 ± 0.00
2 0.09 ± 0.01 0.92 ± 0.01
3 0.10 ± 0.00 0.91 ± 0.00
4 0.09 ± 0.01 0.92 ± 0.00
5 0.09 ± 0.01 0.92 ± 0.00
6 0.09 ± 0.00 0.92 ± 0.00
7 0.08 ± 0.01 0.93 ± 0.01
8 0.08 ± 0.00 0.93 ± 0.00
9 0.08 ± 0.01 0.93 ± 0.01

RGZ BYOL 0 0.06 ± 0.00 0.95 ± 0.00
1 0.06 ± 0.01 0.94 ± 0.00
2 0.06 ± 0.00 0.95 ± 0.00
3 0.06 ± 0.00 0.94 ± 0.00
4 0.07 ± 0.01 0.94 ± 0.01
5 0.07 ± 0.01 0.94 ± 0.01
6 0.06 ± 0.00 0.95 ± 0.00
7 0.06 ± 0.01 0.94 ± 0.01
8 0.06 ± 0.00 0.95 ± 0.00
9 0.06 ± 0.00 0.94 ± 0.00

MGCLS Resnet 0 0.08 ± 0.01 0.93 ± 0.01
1 0.07 ± 0.01 0.94 ± 0.01
2 0.09 ± 0.01 0.92 ± 0.01
3 0.09 ± 0.00 0.92 ± 0.00
4 0.08 ± 0.01 0.93 ± 0.01
5 0.07 ± 0.00 0.94 ± 0.00
6 0.06 ± 0.01 0.94 ± 0.01
7 0.08 ± 0.00 0.93 ± 0.00
8 0.08 ± 0.01 0.92 ± 0.01
9 0.07 ± 0.00 0.94 ± 0.00

Notes. Performance of each backbone on binary morphology classification with MiraBest Confident, as the number of fine-tuned layers in the ResNet increases.

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.