Table 2

Training and testing results for Task1.

Input 6800 image (ACA & N) (5440 image for training & 1360 for testing) (balanced training samples 4836 image)
Epoch Train loss Train accuracy Test loss Test accuracy Test precision Test recall Test F1-score Training time
1 0.3069 96.86% 0.0255 99.85% 100% 100% 100% 51s
2 0.0192 99.77% 0.0215 99.41% 99.00% 99% 99.00% 51s
3 0.0106 100.00% 0.0127 99.93% 100% 100% 100% 52s
4 0.0410 99.05% 0.0316 99.78% 100% 100% 100% 52s
5 0.0263 99.57% 0.0144 100% 100% 100% 100% 53s
6 0.0147 99.88% 0.0103 100% 100% 100% 100% 53s
7 0.0097 100.00% 0.0097 100% 100% 100% 100% 53s
8 0.0100 100.00% 0.0096 100% 100% 100% 100% 53s
9 0.0089 100.00% 0.0090 100% 100% 100% 100% 53s
10 0.0089 100.00% 0.0092 100% 100% 100% 100% 53s
11 0.0090 100.00% 0.0093 100% 100% 100% 100% 53s
12 0.0086 100.00% 0.0089 100% 100% 100% 100% 53s
13 0.0086 100.00% 0.0089 100% 100% 100% 100% 53s
14 0.0087 100.00% 0.0089 100% 100% 100% 100% 54s
15 0.0098 99.98% 0.0089 100% 100% 100% 100% 54s

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.