CrevasseSeg: A Label-Efficient UAV Crevasse Segmentation Framework

Steven Wallace; William D. Harcourt; Richard Hann; Aiden Durrant; Somayajulu Sripada; Georgios Leontidis

cs.LGAug 16, 2026

CrevasseSeg: A Label-Efficient UAV Crevasse Segmentation Framework

Authors: Steven Wallace, William D. Harcourt, Richard Hann, Aiden Durrant, Somayajulu Sripada, Georgios Leontidis

Organizations: School of Natural and Computing Sciences, University of Aberdeen, UK · Interdisciplinary Institute, University of Aberdeen, UK · School of Geosciences, University of Aberdeen, UK · Department of Engineering Cybernetics, Norwegian University of Science and Technology, Norway ·

Abstract

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.

Figures & tables

Split	#Tiles	Labelled	Role
Train	1,938	No	SSL fine-tuning
Val	24	Yes	Fit readout / tune
Test	176	Yes	Evaluation

Table 1: CrevasseSeg split. Labels are available only for the validation and test sets.

Figure 1: Overview of the DINOv3-O-Net architecture. The frozen DINOv3 backbone extracts multi-scale features, which are passed through per-block adaptor layers (DINOv3 Adaptor) and a Feature Alignment/Pyramid Module (FAPM) before being progressively fused by a stack of decoders to produce the final binary crevasse segmentation mask output from XGBoost.

Methods	mDSC	mIoU
DINOv3-ViT-L-O-Net-BYOL-JSD	50.55	34.98
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD	48.27	33.87
O-Net-BYOL-JSD	71.79	57.07
O-Net-Plus-Plus-BYOL-JSD	71.47	57.04
DINOv3-ViT-L-O-Net-BYOL	48.60	34.12
DINOv3-ViT-L-Sat-O-Net-BYOL	48.27	33.87

Table 2: Linear-probe evaluation on CrevasseSeg test (foreground mDSC, mIoU) for five SSL objectives

\times

three backbones. DINOv3 rows are weak and several collapse to the majority-class score.

Methods	mDSC	mIoU
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD	75.33	61.82
DINOv3-ViT-L-O-Net-BYOL-JSD	74.23	60.31
O-Net-BYOL-JSD	73.15	59.25
O-Net-Plus-Plus-BYOL-JSD	73.35	59.58
DINOv3-ViT-L-Sat-O-Net-BYOL	75.68	62.43
DINOv3-ViT-L-O-Net-BYOL	74.23	60.29

Table 3: XGBoost (non-linear) readout on CrevasseSeg test. The ranking inverts: satellite-pretrained DINOv3-O-Net becomes strongest. Best per objective in bold .

Method	mDSC	mIoU
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD (ours)	75.33	61.82
KNN	70.96	56.54
Random Forest	71.58	57.50
XGBoost	71.43	57.12
FNN	71.83	57.55
LSTM	71.25	56.95

Table 4: CrevasseSeg test: our label-efficient pipeline vs. standard baselines fit on the same 24 labels. Best in bold .

Method	mDSC	mIoU
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD-Ensemble (ours)	77.30	64.54
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD (ours)	75.33	61.82
PriMaPs-Em-DINO-ViT-B	75.00	61.11
EAGLE-DINO-ViT-S	77.54	64.65
HP-ViT-S	74.85	60.95
STEGO-DINO-ViT-B	76.43	63.00

Table 5: CrevasseSeg test: our label-efficient pipeline vs. standard deep learning baselines non-linear evaluation. Best in bold .

Method	mDSC	mIoU
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD-Ensemble (ours)	73.64	59.88
DINOv3-ViT-L-Sat-O-Net-BYOL-JSD (ours)	48.27	33.87
PriMaPs-Em-DINO-ViT-B	74.88	60.96
EAGLE-DINO-ViT-S	75.54	62.05
HP-DINO-ViT-S	74.85	60.95
STEGO-DINO-ViT-B	75.84	62.41

Table 6: CrevasseSeg test: our label-efficient pipeline vs. standard deep learning baselines linear evaluation. Best in bold .

Panel of Figure 2 — Figure 2: Comparison between the input tile, ground truth mask, and masks from four evaluation methods, displaying the majority-class collapse under linear evaluation for DINOv3-ViT-L-Sat-O-Net-BYOL-JSD and not the other methods.

Decoder Block	mDSC	mIoU
Block 1	0.7026	0.5586
Block 2	0.7291	0.5886
Block 3	0.4827	0.3387

Table 7: Layer-wise linear probing results for DINOv3-ViT-L-Sat-O-Net-BYOL-JSD decoder blocks, evaluated on upsampled features against pixel-level labels.

Panel of Figure 3 — Figure 3: Two-dimensional UMAP projections of the frozen per-pixel features (BYOL–JSD objective) for the four backbones, coloured by class (class 0: ice/background, $n{=}13{,}168$ ; class 1: crevasse/foreground, $n{=}6{,}832$ ). The DINOv3 backbones (a, b) fragment pixels into many small clusters in which the two classes are locally interleaved, whereas the convolutional backbones (c, d) embed pixels onto a single connected manifold whose ends are dominated by different classes.

Remote Sensing Image Segmentation

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

See all 220 papers

Cryo-Bench: Benchmarking Foundation Models for Cryosphere Mapping
Mar 2, 2026Saurabh Kaushik, Lalit Maurya, Beth Tellman +4Remote Sensing Image Segmentation Geospatial Foundation Models
University of Wisconsin–Madison, Madison, WI, USA · School of Computing and Mathematics, Manchester Metropolitan University, Manchester, M1 5GD, UK · Department of Computer Science and Engineering, Tezpur University, Assam 784028, India +3
CryoNet: A Deep Learning Framework for Multi-Modal Debris-Covered Glacier Mapping. A Case Study of the Poiqu Basin, Central Himalaya
May 19, 2026Farzaneh Barzegar, Tobias Bolch, Norbert Kuehtreiber +1Remote Sensing Image Segmentation Remote Sensing
Institute of Geodesy, Graz University of Technology, Graz, Austria · Engineering Department, University of Sannio, Benevento, Italy
Evaluating Zone-Guided Front Extraction for Glacier Calving-Front Delineation in SAR Imagery
Oct 2, 2026Chhaya Kulkarni, Emam HossainRemote Sensing Image Segmentation Semantic Segmentation
Department of Computer and Information Sciences Towson University · Department of Information Systems University of Maryland, Baltimore County

Mar 2, 2026J/K move · Enter open · S save

Cryo-Bench: Benchmarking Foundation Models for Cryosphere Mapping

Saurabh Kaushik, Lalit Maurya, Beth Tellman, Swalpa Kumar Roy+3

University of Wisconsin–Madison, Madison, WI, USA · School of Computing and Mathematics, Manchester Metropolitan University, Manchester, M1 5GD, UK · Department of Computer Science and Engineering, Tezpur University, Assam 784028, India+3

Remote Sensing Image Segmentation Geospatial Foundation Models Remote Sensing Few-Shot Semantic Segmentation Remote Sensing Benchmarks Semantic Segmentation

Open paper AskPDF arXiv

Geo-Foundation Models (GFMs) have been evaluated across diverse Earth observation tasks and domains, showing strong potential to produce reliable maps even with sparse labels. However, systematic benchmarking of GFMs for Cryosphere applications remains limited, primarily because suitable evaluation datasets are scarce. We address this gap by introducing Cryo-Bench, a benchmark comprising six semantic segmentation datasets covering five cryospheric components: supraglacial debris, glacial lakes under two sensing configurations, sea ice, calving fronts and Antarctic ice-shelf extent. The benchmark includes multispectral, RGB, and synthetic aperture radar observations from regions underrepresented in existing pretraining archives. We evaluate thirteen GFMs alongside U-Net and Vision Transformer baselines trained from scratch under a unified evaluation protocol. With frozen encoders, the U-Net achieves the highest six-dataset average mean intersection over union (mIoU) of 69.31%, exceeding TerraMind (67.86%) by 1.45 points. The paired difference has a 95% confidence interval of [+0.94, +1.98], indicating that the U-Net's lead is statistically significant. In contrast, learning-rate optimization substantially improves fine-tuning performance: DOFA reaches 93.97% mIoU on the RGB glacial lake task, ranking the U-Net fourth, while Scale-MAE and GFM-Swin surpass U-Net on calving fronts. Averaged across all six datasets, four GFMs, TerraMind, GFM-Swin, DOFA, and Scale-MAE, exceed the U-Net baseline (69.31%). In the few-shot setting, five GFMs, DOFA, RemoteCLIP, TerraMind, GFM-Swin, and Scale-MAE, likewise outperform U-Net; averaged across all thirteen GFMs, retention is 92.5% of full-label accuracy compared with 86.1% for U-Net.

Abstract

Remote Sensing Image Segmentation

Explore similar work

Cryo-Bench: Benchmarking Foundation Models for Cryosphere Mapping

CryoNet: A Deep Learning Framework for Multi-Modal Debris-Covered Glacier Mapping. A Case Study of the Poiqu Basin, Central Himalaya

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