Curriculum-guided Change Detection Training: Toward Accurate Serac Fall Monitoring

2026-06-26

AI Summary

This work addresses the limitations of conventional change detection methods that rely on uniform sampling, which assumes equal contribution from all samples and is thus susceptible to noisy gradients, compromising representation robustness. For the first time, curriculum learning is introduced into the change detection domain through a novel difficulty-aware training framework. The proposed approach leverages solar angle gap (SAG) and structural similarity (SSIM) to assess sample difficulty, enabling progressive learning from easy to hard instances. This framework is grounded in clear physical principles and is orthogonal to existing models, offering strong generality. Experiments on the SeracFallDet benchmark demonstrate that the method significantly outperforms uniform sampling strategies, substantially enhancing both pixel-level and object-level change detection accuracy and robustness.

Abstract

Change Detection (CD) aims to identify semantic or structural changes from nearly registered multi-temporal images. While recent advances in training methodologies have largely focused on semi-supervised learning and consistency regularization, alternative training paradigms remain underexplored. In particular, most deep CD methods rely on uniform sampling during training, implicitly assuming that all training samples contribute equally to the optimization process. However, such naive sampling can introduce noisy gradients and hinder robust representation learning. To address this limitation, we propose a curriculum learning framework tailored for change detection. Our approach investigates two complementary difficulty measures: the Solar Angular Gap (SAG), a physically grounded proxy for acquisition-condition variability, and the Structural Similarity Index Measure (SSIM), which evaluates appearance similarity between image pairs. Based on these criteria, the framework progressively introduces challenging samples during training, enabling models to learn robust representations in a coarse-to-fine manner. We evaluate our method on the challenging SeracFallDet benchmark, where results demonstrate consistent improvements of the proposed approach over standard uniform-sampling strategies for both pixel-based and object-based approaches. These results highlight the potential of curriculum learning to improve robustness in deep change detection. Importantly, our training framework is orthogonal to existing CD architectures, making it readily applicable to a broad range of methods.

Problem

• Change Detection
• Curriculum Learning
• Training Sampling
• Robust Representation
• Multi-temporal Images

Innovation

• Curriculum Learning
• Change Detection
• Solar Angular Gap
• SSIM
• Robust Representation Learning