Partial Label Learning with Focal Loss for Sea Ice Classification Based on Ice Charts

Behzad Vahedi,Benjamin Lucas,Farnoush Banaei-Kashani,Andrew P. Barrett,Walter N. Meier,Siri Jodha Khalsa,Morteza Karimzadeh

DOI: https://doi.org/10.1109/JSTARS.2024.3413003

2024-06-10

Abstract:Sea ice, crucial to the Arctic and Earth's climate, requires consistent monitoring and high-resolution mapping. Manual sea ice mapping, however, is time-consuming and subjective, prompting the need for automated deep learning-based classification approaches. However, training these algorithms is challenging because expert-generated ice charts, commonly used as training data, do not map single ice types but instead map polygons with multiple ice types. Moreover, the distribution of various ice types in these charts is frequently imbalanced, resulting in a performance bias towards the dominant class. In this paper, we present a novel GeoAI approach to training sea ice classification by formalizing it as a partial label learning task with explicit confidence scores to address multiple labels and class imbalance. We treat the polygon-level labels as candidate partial labels, assign the corresponding ice concentrations as confidence scores to each candidate label, and integrate them with focal loss to train a Convolutional Neural Network (CNN). Our proposed approach leads to enhanced performance for sea ice classification in Sentinel-1 dual-polarized SAR images, improving classification accuracy (from 87% to 92%) and weighted average F-1 score (from 90% to 93%) compared to the conventional training approach of using one-hot encoded labels and Categorical Cross-Entropy loss. It also improves the F-1 score in 4 out of the 6 sea ice classes.

Computer Vision and Pattern Recognition,Image and Video Processing

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are two major challenges in automated sea - ice classification on training data: the multi - label problem and the class - imbalance problem. Specifically: 1. **Multi - label problem**: Sea - ice maps generated by experts (as training data) are usually not of a single type of sea - ice, but rather polygon areas containing multiple sea - ice types. The traditional one - hot encoding method cannot effectively handle these mixed labels, resulting in information loss and potential mislabeling. 2. **Class - imbalance problem**: The distribution of different types of sea - ice in the data is usually unbalanced, and some types of sea - ice occur far less frequently than others. This will cause the model to be biased towards frequently occurring classes during the training process, and have poor performance on rare classes. To solve these problems, the author proposes a new GeoAI method, formulating the sea - ice classification task as a Partial Label Learning (PLL) problem and introducing Focal Loss to deal with class - imbalance. Specific methods include: - **Partial Label Learning (PLL)**: Consider polygon - level labels as candidate partial labels and assign confidence scores according to the concentration of each sea - ice type. These confidence scores are integrated into the training process to better reflect the existence probability of each sea - ice type. - **Focal Loss**: By adjusting the modulation factor and the focusing parameter in the loss function, reduce the influence of common classes and increase the attention to rare classes, thereby improving the overall classification performance. Through this method, the author has successfully improved the accuracy of sea - ice classification based on Sentinel - 1 dual - polarization SAR images from 87% to 92%, and the weighted average F1 - score from 90% to 93%. In addition, the F1 - score has also been improved in 4 out of 6 sea - ice categories. In summary, this paper aims to overcome the multi - label and class - imbalance problems in automated sea - ice classification through innovative PLL and Focal Loss methods, thereby improving classification accuracy and reliability.

Partial Label Learning with Focal Loss for Sea Ice Classification Based on Ice Charts

Region-level labels in ice charts can produce pixel-level segmentation for Sea Ice types

Sea Ice Image Classification Based on Heterogeneous Data Fusion and Deep Learning

Enhancing sea ice segmentation in Sentinel-1 images with atrous convolutions

A Multi-Task Network with Dynamic Segmentation for Sea Ice Classification in Arctic Shipping Route Optimization

Sea Ice Classification of SAR Imagery Based on Convolution Neural Networks

Deep Semisupervised Teacher–Student Model Based on Label Propagation for Sea Ice Classification

Toward Polar Sea-Ice Classification using Color-based Segmentation and Auto-labeling of Sentinel-2 Imagery to Train an Efficient Deep Learning Model

A Parallel Workflow for Polar Sea-Ice Classification using Auto-labeling of Sentinel-2 Imagery

Data Augmentation for SAR Sea Ice and Water Classification Based on Per-Class Backscatter Variation With Incidence Angle

Fine-Scale Sea Ice Segmentation for High-Resolution Satellite Imagery with Weakly-Supervised CNNs

Deep Learning-Based Seasonal Forecast of Sea Ice Considering Atmospheric Conditions

The Identification of Ice Floes and Calculation of Sea Ice Concentration Based on a Deep Learning Method

Comparison of Cross-Entropy, Dice, and Focal Loss for Sea Ice Type Segmentation

Geo-knowledge-informed Deep Learning for Auto-identification of Supraglacial Lakes on the Greenland Ice Sheet from Satellite Imagery.

DF-UHRNet: A Modified CNN-Based Deep Learning Method for Automatic Sea Ice Classification from Sentinel-1A/B SAR Images

IceGCN: An Interactive Sea Ice Classification Pipeline for SAR Imagery Based on Graph Convolutional Network

MMSeaIce: a collection of techniques for improving sea ice mapping with a multi-task model

Decoding the Partial Pretrained Networks for Sea-Ice Segmentation of 2021 Gaofen Challenge

Sea Ice Detection from GNSS-R Data Based on Local Linear Embedding

Deep Learning Based Sea Ice Classification with Gaofen-3 Fully Polarimetric SAR Data