Evaluating Zone-Guided Front Extraction for Glacier Calving-Front Delineation in SAR Imagery

Chhaya Kulkarni; Emam Hossain

cs.CVOct 2, 2026

Evaluating Zone-Guided Front Extraction for Glacier Calving-Front Delineation in SAR Imagery

Authors: Chhaya Kulkarni, Emam Hossain

Organizations: Department of Computer and Information Sciences Towson University · Department of Information Systems University of Maryland, Baltimore County

Abstract

Automatic calving-front delineation from synthetic aperture radar imagery is challenging because the front is a thin and often ambiguous boundary between glacier ice, ocean, and surrounding rock or terrain. The CAlving Fronts and where to Find thEm (CaFFe) dataset provides both binary calving-front masks and broader semantic zone masks, making it possible to study whether zone-level supervision can support front recovery. In this paper, we compare direct front prediction with zone-guided front extraction using U-Net, DeepLabV3+, and SegFormer-B0 under the same bounding-box-cropped CaFFe setting. In the direct setting, models predict the binary calving-front mask. In the zone-guided setting, the model first predicts four semantic zone classes, and the front is then extracted from the predicted glacier-ocean boundary. We evaluate both zone-level and front-level performance, include a ground-truth-zone boundary check, and examine lightweight test-time adaptation on sensor-specific and glacier-specific subsets. The results show that zone labels contain useful front-boundary information: extracting the front from ground-truth zones gives the lowest mean distance error. However, fronts extracted from model-predicted zones remain weak, even when zone segmentation scores are moderate. Test-time adaptation also does not consistently improve zone-guided front recovery. These results indicate that zone segmentation performance should not be treated as a substitute for front-level evaluation and that effective use of zone labels may require boundary-aware training, label fusion, or explicit front supervision.

Figures & tables

Fig. 1 : Overview of the experimental pipeline. The same bounding-box-cropped SAR inputs are evaluated under two supervision settings: direct front prediction and zone-guided front extraction. In the zone-guided branch, the calving front is recovered from the predicted glacier–ocean boundary. Subset analyses on the Sentinel-1 and Mapple subsets apply flip and BatchNorm test-time adaptation. Colors in the zone masks denote the four semantic classes (ocean, rock outcrop, glacier, no-information).

Model	Setting	Zone F1 $\uparrow$	Front F1 $\uparrow$	MDE $\downarrow$
U-Net	Direct front	–	0.0510	79.32
U-Net	Zone-guided	0.5814	0.0055	64.65
DeepLabV3+	Direct front	–	0.0427	65.47
DeepLabV3+	Zone-guided	0.3642	0.0060	90.06
SegFormer-B0	Direct front	–	0.0939	35.79
SegFormer-B0	Zone-guided	0.6808	0.0058	66.37

TABLE I : Direct front prediction vs. zone-guided front extraction on the CaFFe test set. The ground-truth-zone row (italic) is a diagnostic upper reference, not a trained model.

Model	Target	Set	Strategy	Zone F1 $\uparrow$	Front F1 $\uparrow$	MDE $\downarrow$
Seg-B0	Front	S1	None	–	0.1109	43.82
		S1	Flip	–	0.1150	38.13
		Mapple	None	–	0.1105	36.15
		Mapple	Flip	–	0.1054	33.98
	Zone	S1	None	0.5134	0.0018	97.67
		S1	Flip	0.5174	0.0022	102.83

TABLE II : Test-time adaptation on the worst-case Sentinel-1 (S1) and Mapple subsets. “Front” denotes models trained on binary front masks; “Zone” denotes models trained on zone masks with the front extracted from the predicted glacier-ocean boundary.

Remote Sensing Image Segmentation

Remote sensing image segmentation labels land cover and objects in satellite, aerial, radar, and multispectral imagery.

See all 220 papers

CrevasseSeg: A Label-Efficient UAV Crevasse Segmentation Framework
Aug 16, 2026Steven Wallace, William D. Harcourt, Richard Hann +3Remote Sensing Image Segmentation Self-Supervised Learning
School of Natural and Computing Sciences, University of Aberdeen, UK · Interdisciplinary Institute, University of Aberdeen, UK · School of Geosciences, University of Aberdeen, UK +3
Preserving the Manifold: Frozen Expert Alignment for Label-Efficient SAR Scene Understanding
Oct 18, 2026Eli A. Underwood, I. Taylor McKechnie, Leslie M. CollinsRemote Sensing VLMs Cross-Modal Alignment
Duke University
GlacierCastAI: Predicting Glacier Retreat from Multi-Modal Satellite Imagery and Climate Signals
Jul 5, 2026Arunkumar RamachandranClimate Science Remote Sensing
Independent Researcher

Aug 16, 2026J/K move · Enter open · S save

CrevasseSeg: A Label-Efficient UAV Crevasse Segmentation Framework

Steven Wallace, William D. Harcourt, Richard Hann, Aiden Durrant+2

School of Natural and Computing Sciences, University of Aberdeen, UK · Interdisciplinary Institute, University of Aberdeen, UK · School of Geosciences, University of Aberdeen, UK+3

Remote Sensing Image Segmentation Self-Supervised Learning UAV Remote Sensing Remote Sensing Benchmarks

Open paper AskPDF arXiv

Crevasse mapping from uncrewed aerial vehicle (UAV) imagery matters for glaciological research and for field safety in glaciated terrain. Yet, pixel-level annotation of glacier surfaces is costly and requires domain experts. We introduce CrevasseSeg, a framework for binary segmentation over the terminus of Borebreen, Svalbard, comprising 1,938 unlabelled UAV orthomosaic tiles for self-supervised/unsupervised fine-tuning, 24 labelled tiles for validation and 176 labelled tiles for testing. Using CrevasseSeg, we benchmark five self-supervised objectives -- BYOL, a Jensen-Shannon Divergence (JSD) objective, Barlow-Twins, VICReg, and a combined BYOL-JSD objective -- across three architectures: O-Net, O-Net++, and a DINOv3-initialised O-Net. Each configuration is evaluated under two frozen-feature readouts that differ only in the form of their decision boundary: a linear probe and a non-linear XGBoost classifier fit only on the 24 labelled validation images. Our central finding is a consistent inversion between the two readouts: DINOv3 features are the weakest under linear probing but the strongest under a non-linear readout. A UMAP analysis of the learned feature space shows that DINOv3 fragments pixels into many small clusters in which the classes are locally interleaved, whereas the convolutional architectures (O-Net and O-Net++) embed them onto a single class-sorted manifold. Satellite-pretrained DINOv3 improves over natural-image initialisation across objectives, and our label-efficient DINOv3-ViT-L-Sat-O-Net-BYOL-JSD pipeline reaches 75.33 mDSC / 61.28 mIoU, outperforming standard machine learning baselines fit on the same 24 labelled images with the RGB pixel values used as features. We release CrevasseSeg to support label-efficient segmentation research in remote sensing.

Abstract

Remote Sensing Image Segmentation

Explore similar work

CrevasseSeg: A Label-Efficient UAV Crevasse Segmentation Framework

Preserving the Manifold: Frozen Expert Alignment for Label-Efficient SAR Scene Understanding

GlacierCastAI: Predicting Glacier Retreat from Multi-Modal Satellite Imagery and Climate Signals