Label-frugal satellite image change detection with generative virtual exemplar learning

2025-10-08

AI Summary

To address the high annotation cost and scarcity of labeled data in satellite image change detection, this paper proposes a generative virtual exemplar learning framework. The method jointly models sample representativeness, diversity, and ambiguity by integrating an invertible graph convolutional network with adversarial loss, enabling automatic importance assessment and high-quality virtual sample generation from unlabeled data to guide critical sample selection in active learning. Unlike conventional active learning that selects only real samples, our approach compensates for annotation coverage gaps via controllable generation, substantially reducing manual labeling effort. Experiments on multiple public change detection benchmarks demonstrate that the method achieves superior performance using only 30% of the labeled data—outperforming fully supervised baselines with mIoU gains of 4.2–6.8 percentage points—validating its dual advantages in labeling efficiency and detection accuracy.

Technology Category

• Machine Learning: Active Learning
• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Web Mining and Content Analysis: Web data generation and simulation

Abstract

Change detection is a major task in remote sensing which consists in finding all the occurrences of changes in multi-temporal satellite or aerial images. The success of existing methods, and particularly deep learning ones, is tributary to the availability of hand-labeled training data that capture the acquisition conditions and the subjectivity of the user (oracle). In this paper, we devise a novel change detection algorithm, based on active learning. The main contribution of our work resides in a new model that measures how important is each unlabeled sample, and provides an oracle with only the most critical samples (also referred to as virtual exemplars) for further labeling. These exemplars are generated, using an invertible graph convnet, as the optimum of an adversarial loss that (i) measures representativity, diversity and ambiguity of the data, and thereby (ii) challenges (the most) the current change detection criteria, leading to a better re-estimate of these criteria in the subsequent iterations of active learning. Extensive experiments show the positive impact of our label-efficient learning model against comparative methods.

Problem

• Reduces labeling effort for satellite change detection
• Generates critical virtual exemplars for active learning
• Improves change detection with adversarial loss optimization

Innovation

• Active learning selects critical unlabeled samples
• Generative model creates virtual exemplars via adversarial loss
• Invertible graph convnet optimizes data representativity and diversity