River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs
To obtain reliable water segmentations from image data for real-time monitoring of river water levels, a comparison of 32 convolutional neural networks was performed. They were trained on a new river water segmentation dataset consisting of 1128 images. To prevent overfitting, two methods using offl...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2023-05-01
|
| Series: | International Journal of Applied Earth Observations and Geoinformation |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1569843223001279 |
| _version_ | 1797828183050944512 |
|---|---|
| author | Franz Wagner Anette Eltner Hans-Gerd Maas |
| author_facet | Franz Wagner Anette Eltner Hans-Gerd Maas |
| author_sort | Franz Wagner |
| collection | DOAJ |
| description | To obtain reliable water segmentations from image data for real-time monitoring of river water levels, a comparison of 32 convolutional neural networks was performed. They were trained on a new river water segmentation dataset consisting of 1128 images. To prevent overfitting, two methods using offline and online augmentation were developed to improve the variance. It was found that offline augmentation is superior on fewer data, while online augmentation is advantageous for a larger dataset (such as Cityscapes).The network comparison showed that U-Net performs best on the water segmentation dataset when using an ResNeXt 50 encoding network pre-trained on ImageNet. It achieves an intersection over union (IoU) of 0.91 without augmentation, 0.98 with offline augmentation and 0.93 with the online augmentation method. The authors have applied the algorithms for online and offline augmentation to Cityscapes to verify the applicability of the strategies to other datasets. The mean IoU is 0.86 without augmentation, 0.86 with offline augmentation and 0.87 with online augmentation. Only online augmentation could prevent overfitting on Cityscapes. |
| first_indexed | 2024-04-09T13:00:20Z |
| format | Article |
| id | doaj.art-4cc473fbb3a44f81a87525c767d88a54 |
| institution | Directory Open Access Journal |
| issn | 1569-8432 |
| language | English |
| last_indexed | 2024-04-09T13:00:20Z |
| publishDate | 2023-05-01 |
| publisher | Elsevier |
| record_format | Article |
| series | International Journal of Applied Earth Observations and Geoinformation |
| spelling | doaj.art-4cc473fbb3a44f81a87525c767d88a542023-05-13T04:24:32ZengElsevierInternational Journal of Applied Earth Observations and Geoinformation1569-84322023-05-01119103305River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNsFranz Wagner0Anette Eltner1Hans-Gerd Maas2Chair of Photogrammetry, TU Dresden, Dresden, Germany; Corresponding author.Junior Professorship in Geosensor Systems, TU Dresden, Dresden, GermanyChair of Photogrammetry, TU Dresden, Dresden, GermanyTo obtain reliable water segmentations from image data for real-time monitoring of river water levels, a comparison of 32 convolutional neural networks was performed. They were trained on a new river water segmentation dataset consisting of 1128 images. To prevent overfitting, two methods using offline and online augmentation were developed to improve the variance. It was found that offline augmentation is superior on fewer data, while online augmentation is advantageous for a larger dataset (such as Cityscapes).The network comparison showed that U-Net performs best on the water segmentation dataset when using an ResNeXt 50 encoding network pre-trained on ImageNet. It achieves an intersection over union (IoU) of 0.91 without augmentation, 0.98 with offline augmentation and 0.93 with the online augmentation method. The authors have applied the algorithms for online and offline augmentation to Cityscapes to verify the applicability of the strategies to other datasets. The mean IoU is 0.86 without augmentation, 0.86 with offline augmentation and 0.87 with online augmentation. Only online augmentation could prevent overfitting on Cityscapes.http://www.sciencedirect.com/science/article/pii/S1569843223001279Deep learningSegmentationWater segmentationAugmentationCNN comparison |
| spellingShingle | Franz Wagner Anette Eltner Hans-Gerd Maas River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs International Journal of Applied Earth Observations and Geoinformation Deep learning Segmentation Water segmentation Augmentation CNN comparison |
| title | River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs |
| title_full | River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs |
| title_fullStr | River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs |
| title_full_unstemmed | River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs |
| title_short | River water segmentation in surveillance camera images: A comparative study of offline and online augmentation using 32 CNNs |
| title_sort | river water segmentation in surveillance camera images a comparative study of offline and online augmentation using 32 cnns |
| topic | Deep learning Segmentation Water segmentation Augmentation CNN comparison |
| url | http://www.sciencedirect.com/science/article/pii/S1569843223001279 |
| work_keys_str_mv | AT franzwagner riverwatersegmentationinsurveillancecameraimagesacomparativestudyofofflineandonlineaugmentationusing32cnns AT anetteeltner riverwatersegmentationinsurveillancecameraimagesacomparativestudyofofflineandonlineaugmentationusing32cnns AT hansgerdmaas riverwatersegmentationinsurveillancecameraimagesacomparativestudyofofflineandonlineaugmentationusing32cnns |