Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.18758987 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>This dataset contains flood monitoring camera images collected from four gauging <br>stations in the United Kingdom: Tewkesbury, DiglisLock, Strensham, and Evesham.</p> <p>The dataset is used to support water body segmentation tasks based on a <br>personalized one-shot segmentation approach (PerSAM with KMeans prompt selection).</p> <p>Each site includes:<br>- Images/water/: Raw camera images (.jpg) captured at regular intervals<br>- Annotations/water/: Ground-truth binary water body masks (.png), <br> where water pixels are labeled as 128<br>- neg_Annotations/refine_mask/: Refined background region masks (.png), <br> where background pixels are labeled as 128, used for negative prompt extraction</p> <p>The dataset covers a range of flood conditions and water level variations, <br>enabling evaluation of segmentation robustness across diverse environmental settings.<br><br>In addition to the image dataset, this record includes pre-trained model weights <br>for a 5-fold cross-validation UNet-ResNet50 segmentation model trained on each <br>of the four sites. The weights are provided as 20 checkpoint files <br>(5 folds × 4 sites), named in the format {Site}_fold{N}_best.pt. <br>These weights correspond to the best-performing checkpoint (lowest validation <br>loss) for each fold and can be used to reproduce the segmentation results <br>reported in the associated paper.</p> <p>Training framework: RIWA Segmentation (Wagner et al.), available at <a href="https://gitlab.com/fra-wa/pytorch_segmentation">https://gitlab.com/fra-wa/pytorch_segmentation</a><br>Associated code repository: <a href="https://github.com/hupi11/flood-water-segmentation">https://github.com/hupi11/flood-water-segmentation</a></p>