Zero-OVCD: Bridging Training-Free Foundation Models and Pseudo-Label Learning for Open-Vocabulary Change Detection

2026-08-12

AI Summary

This work addresses the challenges of inaccurate candidate masks, semantic ambiguity, and error propagation in open-vocabulary change detection by proposing Zero-OVCD, a two-stage framework that operates without pixel-level annotations in the target domain. In the first stage, high-quality change pseudo-labels are generated through a fusion of multiple strategies; in the second stage, a detector is trained on these pseudo-labels with an integrated noise suppression mechanism. The approach innovatively combines zero-shot inference from foundation models with noise-robust pseudo-label learning, enhancing robustness via complementary mask refinement, multi-scale semantic similarity fusion, response-guided correction, checkpoint voting, and selection of highly consistent samples. The method achieves F1 scores of 88.65%, 88.85%, and 57.96% on LEVIR-CD, WHU-CD, and S2Looking, respectively, and attains a macro-average F1 of 50.92% across six tasks on the SECOND benchmark.

Abstract

Open-vocabulary change detection (OVCD) enables the identification of user-specified land-cover changes in bitemporal remote sensing images, but existing training-free pipelines remain vulnerable to inaccurate candidate masks, ambiguous semantic assignments, and accumulated inference errors. To address these issues, we propose Zero-OVCD, a two-stage framework that requires no pixel-level annotations from the target domain. In the first stage, high-quality change pseudo-labels are generated through complementary candidate-mask refinement, multiscale semantic similarity fusion with margin-based reliability filtering, and response-guided mask correction and completion. These components jointly suppress noisy candidates, enhance mask-level semantic discrimination, and recover missed change regions. In the second stage, a change detector is trained using the generated pseudo-labels, while checkpoint voting and high-agreement sample selection are introduced to mitigate residual pseudo-label noise. On LEVIR-CD, WHU-CD, and S2Looking, Stage I achieves F1 scores of 86.25%, 85.82%, and 50.48%, while Stage II further improves them to 88.65%, 88.85%, and 57.96%, respectively. On SECOND, the macro-average F1 across six category-wise one-vs-rest tasks increases from 47.91% to 50.92%. These results demonstrate that bridging training-free foundation-model inference with noise-aware pseudo-label learning provides an effective solution for open-vocabulary change detection without target-domain pixel-level annotations. Code will be available at https://github.com/1321663019/Zero-OVCD.

Problem

• open-vocabulary change detection
• candidate mask inaccuracy
• ambiguous semantic assignment
• inference error accumulation
• pseudo-label noise

Innovation

• open-vocabulary change detection
• pseudo-label learning
• training-free foundation models
• mask refinement
• noise-aware learning