Front. Earth Sci.Frontiers in Earth ScienceFront. Earth Sci.2296-6463Frontiers Media S.A.133942610.3389/feart.2023.1339426Earth ScienceCorrectionCorrigendum: Extraction of stratigraphic exposures on visible images using a supervised machine learning techniqueNoguchi and Shoji10.3389/feart.2023.1339426NoguchiRina1*ShojiDaigo21Faculty of Science, Niigata University, Niigata, Japan2Institute of Space and Astronautical Science, Japan Aerospace Exploration Agency, Sagamihara, Kanagawa, Japan
Approved by:Frontiers Editorial Office, Frontiers Media SA, Switzerland
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
A Corrigendum on Extraction of stratigraphic exposures on visible images using a supervised machine learning technique by Noguchi R and Shoji D (2023). Front. Earth Sci. 11:1264701. doi: 10.3389/feart.2023.1264701machine learningstratigraphic exposuresvolcanic stratigraphiesautonomous extractionU-NetLinkNetsection-at-acceptanceVolcanology
In the published article, there was an error in the legend for Figure 2 as published. The symbols were attributed to the incorrect IDs. The blue triangle—U-Net fn ResNet18—was described as U-7 but should have been U-4. The green star—U-Net fn ResNet50 - was described as U-13 but should have been U-7. The blue square—U-Net fy ResNet18 was described as U-4 but should have been U-10. The green diamond—U-Net fy ResNet50—was described as U-10 but should have been U-13. The order has been rearranged into the numerical order of the IDs. The correct figure and legend appears below.
Loss, dice loss, and accuracy plots and curves during training for U-Net models with a setting of 50 steps per epoch.
In the published article, there was an error in the legend for Figure 3 as published. The symbols were attributed to the incorrect IDs. The blue triangle—U-Net fn ResNet18—was described as U-8 but should have been U-5. The green star—U-Net fn ResNet50 - was described as U-14 but should have been U-8. The blue square—U-Net fy ResNet18 was described as U-5 but should have been U-11. The green diamond—U-Net fy ResNet50—was described as U-11 but should have been U-14. The order has been rearranged into the numerical order of the IDs. The correct figure and legend appears below.
Loss, dice loss, and accuracy plots and curves during training for U-Net models with a setting of 100 steps per epoch.
In the published article, there was an error in the legend for Figure 4 as published. The symbols were attributed to the incorrect IDs. The blue triangle—U-Net fn ResNet18—was described as U-9 but should have been U-6. The green star—U-Net fn ResNet50—was described as U-15 but should have been U-9. The blue square—U-Net fy ResNet18 was described as U-6 but should have been U-12. The green diamond—U-Net fy ResNet50—was described as U-12 but should have been U-15. The order has been rearranged into the numerical order of the IDs. The correct figure and legend appears below.
Loss, dice loss, and accuracy plots and curves during training for U-Net models with a setting of 150 steps per epoch.
In the published article, there was an error in the legend for Figure 5 as published. The symbols were attributed to the incorrect IDs. The blue triangle–LinkNet fn ResNet18 - was described as L-4 but should have been L-1. The green star—LinkNet fn ResNet50—was described as L-10 but should have been L-4. The blue square—LinkNet fy ResNet18 was described as L-1 but should have been L-7. The green diamond—LinkNet fy ResNet50—was described as L-7 but should have been L-10. The order has been rearranged into the numerical order of the IDs. The correct figure and legend appears below.
Loss, dice loss, and accuracy plots and curves during training for LinkNet models with a setting of 50 steps per epoch.
In the published article, there was an error in the legend for Figure 6 as published. The symbols were attributed to the incorrect IDs. The blue triangle–LinkNet fn ResNet18—was described as L-5 but should have been L-2. The green star—LinkNet fn ResNet50 - was described as L-11 but should have been L-5. The blue square—LinkNet fy ResNet18 was described as L-2 but should have been L-8. The green diamond—LinkNet fy ResNet50—was described as L-8 but should have been L-11. The order has been rearranged into the numerical order of the IDs. The correct figure and legend appears below.
Loss, dice loss, and accuracy plots and curves during training for LinkNet models with a setting of 100 steps per epoch.
In the published article, there was an error in the legend for Figure 7 as published. The symbols were attributed to the incorrect IDs. The blue triangle—LinkNet fn ResNet18—was described as L-6 but should have been L-3. The blue square—LinkNet fy ResNet18 was described as L-3 but should have been L-9. The green star—LinkNet fn ResNet50—was described as L-12 but should have been L-6. The green diamond—LinkNet fy ResNet50—was described as L-9 but should have been L-12. The order has been rearranged into the numerical order of the IDs. The correct figure and legend appears below.
Loss, dice loss, and accuracy plots and curves during training for LinkNet models with a setting of 150 steps per epoch.
In the published article, there was an error in Table 1 as published.
The loss, dice, and binary accuracy at the end of training in each model.
ID
Model
Fine-tuning
Backbone
Step
Training stopped epoch
Training
Validation
Duration (min)
Loss
Dice loss
Accuracy
Loss
Dice loss
Accuracy
Total
Per epoch
U-1
U-Net
No
No
50
58
0.582
0.422
0.785
0.573
0.569
0.646
512
8.83
U-2
100
52
0.568
0.416
0.802
0.582
0.573
0.638
298
5.73
U-3
150
48
0.563
0.411
0.810
0.577
0.573
0.663
282
5.88
U-4
No
Yes (ResNet18)
50
74
0.468
0.290
0.785
0.632
0.509
0.648
90
1.22
U-5
100
62
0.446
0.263
0.791
0.631
0.484
0.659
70
1.13
U-6
150
62
0.428
0.255
0.799
0.644
0.485
0.670
76
1.23
U-7
No
Yes (ResNet50)
50
73
0.450
0.289
0.800
0.591
0.506
0.672
121
1.66
U-8
100
49
0.451
0.273
0.790
0.616
0.493
0.665
78
1.59
U-9
150
44
0.465
0.300
0.781
0.587
0.506
0.671
76
1.73
U-10
Yes
Yes (ResNet18)
50
63
0.322
0.214
0.882
0.584
0.505
0.686
76
1.21
U-11
100
55
0.315
0.220
0.911
0.567
0.496
0.697
61
1.11
U-12
150
43
0.258
0.158
0.906
0.643
0.469
0.693
53
1.23
U-13
Yes
Yes (ResNet50)
50
138
0.254
0.195
0.954
0.507
0.471
0.733
229
1.66
U-14
100
62
0.206
0.139
0.938
0.567
0.442
0.728
99
1.60
U-15
150
63
0.187
0.128
0.947
0.529
0.428
0.748
108
1.71
L-1
LinkNet
No
Yes (ResNet18)
50
28
0.731
0.455
0.531
0.738
0.607
0.509
17
0.61
L-2
100
21
0.772
0.510
0.467
0.680
0.629
0.599
14
0.67
L-3
150
20
0.852
0.567
0.443
0.692
0.667
0.620
15
0.75
L-4
No
Yes (ResNet50)
50
20
0.666
0.412
0.586
0.712
0.583
0.493
30
1.50
L-5
100
108
0.519
0.321
0.742
0.615
0.506
0.646
146
1.35
L-6
150
134
0.508
0.293
0.751
0.645
0.486
0.647
190
1.42
L-7
Yes
Yes (ResNet18)
50
23
0.777
0.351
0.582
1.090
0.535
0.421
14
0.61
L-8
100
20
0.677
0.417
0.593
0.723
0.572
0.514
13
0.65
L-9
150
138
0.393
0.230
0.824
0.587
0.478
0.705
99
0.72
L-10
Yes
Yes (ResNet50)
50
88
0.328
0.206
0.873
0.586
0.483
0.689
126
1.43
L-11
100
104
0.314
0.219
0.898
0.562
0.494
0.697
141
1.36
L-12
150
88
0.303
0.183
0.896
0.678
0.465
0.706
128
1.45
Values for the model ID L-1 and L-7 were incorrect. The corrected Table 1 and its caption appear below.
The authors apologize for these errors and state that this does not change the scientific conclusions of the article in any way. The original article has been updated.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.