Changes in Gaza: DINOv3-Powered Multi-Class Change Detection for Damage Assessment in Conflict Zones

TL;DR

This work tackles fine-grained damage assessment in conflict zones using multi-temporal remote sensing. It introduces a multi-class change detection (MCD) paradigm and a DINOv3-based MC-DiSNet with a Multi-Scale Cross-Attention Difference module, plus a Gaza-change dataset to support humanitarian analysis. Across Gaza-change and external benchmarks (SECOND, Landsat-SCD), the approach achieves state-of-the-art performance and demonstrates strong generalization while maintaining efficiency through LoRA-based fine-tuning and a lightweight decoder. The study advances automated, rapid, and scalable damage assessment in conflict areas, with practical relevance for emergency response and reconstruction planning.

Abstract

Accurately and swiftly assessing damage from conflicts is crucial for humanitarian aid and regional stability. In conflict zones, damaged zones often share similar architectural styles, with damage typically covering small areas and exhibiting blurred boundaries. These characteristics lead to limited data, annotation difficulties, and significant recognition challenges, including high intra-class similarity and ambiguous semantic changes. To address these issues, we introduce a pre-trained DINOv3 model and propose a multi-scale cross-attention difference siamese network (MC-DiSNet). The powerful visual representation capability of the DINOv3 backbone enables robust and rich feature extraction from bi-temporal remote sensing images. The multi-scale cross-attention mechanism allows for precise localization of subtle semantic changes, while the difference siamese structure enhances inter-class feature discrimination, enabling fine-grained semantic change detection. Furthermore, a simple yet powerful lightweight decoder is designed to generate clear detection maps while maintaining high efficiency. We also release a new Gaza-change dataset containing high-resolution satellite image pairs from 2023-2024 with pixel-level semantic change annotations. It is worth emphasizing that our annotations only include semantic pixels of changed areas. We evaluated our method on the Gaza-Change and two classical datasets: the SECOND and Landsat-SCD datasets. Experimental results demonstrate that our proposed approach effectively addresses the MCD task, and its outstanding performance paves the way for practical applications in rapid damage assessment across conflict zones.

Figures (8)

• Figure 1: Evolution of change detection paradigms: (a) Binary Change Detection (BCD), (b) Semantic Change Detection (SCD), and (c) Multi-Class Change Detection (MCD).
• Figure 2: The panoramic remote sensing satellite image of the Gaza Strip.
• Figure 3: Overall architecture of the proposed MC-DiSNet.
• Figure 4: Examples of the proposed Gaza-Change, six distinct colors to encode change categories: red highlights “Building Damage", green marks “New Building", blue indicates “New Camp", yellow denotes "Farmland Damage", purple signals "Greenhouse Damage", whereas cyan represents "New Greenhouse".
• Figure 5: Example results on Gaza-change dataset.