Existing approaches for calving front detection generally work by first performing a pixel-wise segmentation or edge detection, and then extract the actual calving front in a post-processing step. Our goal in this study is to build a model that only needs a single step, and directly outputs the calving front as a polyline.

Following the idea of explicit contour prediction, we have developed a new method called “Charting Outlines by Recurrent Adaptation” (COBRA). It works by combining the idea of Active Contour models with deep learning. First, a 2D CNN backbone derives feature maps from the input imagery. Then, a 1D CNN (Snake Head) iteratively deforms an initial contour until to match the true contour.

These animations show how COBRA iteratively predicts glacier calving fronts:

 

 

 

Preprint available on arxiv: https://arxiv.org/abs/2307.03461

Update: Published article now available on IEEE Explore: https://ieeexplore.ieee.org/document/10195954

If you would like to have a closer look at the implementation details, work with our method, or reproduce our results, you can find all of our code on Github .

Open Fullscreen Map