The Landsat Irish Coastal Segmentation (LICS) Dataset

We introduce the Landsat Irish Coastal Segmentation (LICS) Dataset. Its purpose is to aid the development of deep learning methods for coastal water body segmentation. At the same time, the dataset is used to shed light on modelling challenges that are inherent to Irish meteorology and geology.

To understand more about how this dataset was created, please see this paper: 

O’Sullivan, C., Kashyap, A., Coveney, S., Monteys, X. and Dev, S., 2024. Enhancing coastal water body segmentation with Landsat Irish Coastal Segmentation (LICS) dataset. Remote Sensing Applications: Society and Environment, p.101276. https://arxiv.org/abs/2409.15311

test.zip: 100 test images with the following channels:

• 1-7: spectral bands
• 8: Landsat QA
• 9: rough training annotation
• 10: test mask annotation
• 11: test edge annotation

training.zip: 30,000 training images with the following channels:

• 1-7: spectral bands
• 8: Landsat QA
• 9: rough training annotation

finetune.zip: 100 fine-tuning images with the following channels:

• 1-7: spectral bands 
• 8: Landsat QA
• 9: rough training annotation
• 10: fine-tunning mask annotation

    
swed_finetune.zip: 100 fine-tuning images to accompany SWED

    
model_weights.zip: model weights from paper