Awesome
Coastal Image Labeler
It's a labeling tool!
Labeled images are important for supervised machine learning. The Coastal Image Labeler is focused on easily accommodating multiple users labeling the same images to ensure consensus for (potential ambiguous) discipline-specific labels. This tool allows us to crowdsource the development of a labeled image dataset that is relevant for coastal scientists.
The labeler is deployed here: https://coastalimagelabeler.science/. You can log on and start labeling post Hurricane images collected by NOAA NGS.
You can check out the Coastal Image Labeler documentation for more info:
Code of Conduct
We hope to foster an inclusive and respectful environment surrounding the contribution and discussion of our project. Make sure you understand our Code of Conduct.
Projects using the Labeler:
Papers:
An Active Learning Pipeline to Detect Hurricane Washover in Post-Storm Aerial Images:
Labeling Poststorm Coastal Imagery for Machine Learning: Measurement of Interrater Agreement: https://doi.org/10.1029/2021EA001896
Data Releases:
Labels for Hurricane Florence imagery: 343 images, labeled via consensus, available via figshare.
Labels for Hx Florence, Hx Michael and Hx Isaias imagery:
- v1 (300 images; 2.1k labels):
- v1.1 (900 images; 4.5k labels):
- v1.2: (1500 images; 6.2k labels) :
- v1.3: (1500 images; 6.2k labels; 100 image quadrants (smaller scale) w/ 400 labels):
- v1.4: (1500 images; 6.2k labels; 100 image quadrants (smaller scale) w/ 400 labels; 100 images labeled by noncoastal experts - 400 labels):
- v2.0: (4250 images; 10.2k labels; 100 image quadrants (smaller scale) w/ 400 labels; 100 images labeled by noncoastal experts - 400 labels):
You can merge these labels with post-storm images, which can be downloaded from NOAA NGS or using this nifty python command line downloading tool. We are using these labeled images to detect washover.