Data and Code for “A Hausdorff-Guided Deep Learning Approach for Monitoring the Motion of Rotating Arctic Ice Floes”

Data for the manuscript entitled “A Hausdorff-Guided Deep Learning Approach for Monitoring the Motion of Rotating Arctic Ice Floes.”

This dataset is specifically curated to support the submission and reproducibility of the above-mentioned manuscript. It provides a comprehensive collection of resources covering the full experimental pipeline, including source data, intermediate processing results, comparative experiments, final outputs, and implementation code.

The repository contains both data and code:
Dataset for HDL-IFM
This archive contains all data used in the study and is further organized into five subfolders, which systematically document the workflow from raw data to experimental results:
Source Data
This folder contains 226 multi-temporal Arctic ice floe images used as training data for the deep learning framework proposed in the manuscript.
Ice Floe Data for the Experiment
This folder includes both the original ice floe imagery and the preprocessed data used in the experiments, reflecting the data preparation procedures applied in the study.
Extraction and Matching Results of Traditional Methods for Ice Floe Monitoring
This folder presents the feature extraction and matching results obtained using traditional methods, including SIFT and A-KAZE, which serve as baseline comparisons.
Experimental Results
This folder contains the motion monitoring results of ice floes between consecutive days, as well as vector maps illustrating short-term continuous motion trajectories.
Influence of Ice Floe Rotation on Deep-Model Matching
This folder includes spatial visualization results under controlled rotations (72 rotations with a 5° interval), along with quantitative analyses demonstrating the impact of rotation on matching accuracy.

Code for HDL-IFM
This folder contains the implementation of the proposed Hausdorff-guided deep learning framework (HDL-IFM), including scripts for data preprocessing, model training, feature extraction, matching, and evaluation. The code is organized to facilitate reproducibility of all experiments reported in the manuscript.

This dataset and code repository jointly support the development, validation, and comparative analysis of the proposed framework for detecting feature points and estimating the motion of rotating ice floes under complex Arctic marginal ice zone conditions.