OmniOVCD: Streamlining Open-Vocabulary Change Detection with SAM 3

TL;DR

OmniOVCD addresses open-vocabulary change detection by replacing multi-model pipelines with a single SAM 3–based framework. It introduces SFID, which fuses semantic, instance, and presence heads to produce robust semantic maps and then decouples them into instance masks for bi-temporal matching. The approach yields state-of-the-art IoU across LEVIR-CD, WHU-CD, S2Looking, and SECOND, with improved stability and efficiency thanks to the unified architecture. This work provides a practical, scalable solution for open-world remote sensing analysis and paves the way for further integration of foundation models in CD tasks.

Abstract

Change Detection (CD) is a fundamental task in remote sensing. It monitors the evolution of land cover over time. Based on this, Open-Vocabulary Change Detection (OVCD) introduces a new requirement. It aims to reduce the reliance on predefined categories. Existing training-free OVCD methods mostly use CLIP to identify categories. These methods also need extra models like DINO to extract features. However, combining different models often causes problems in matching features and makes the system unstable. Recently, the Segment Anything Model 3 (SAM 3) is introduced. It integrates segmentation and identification capabilities within one promptable model, which offers new possibilities for the OVCD task. In this paper, we propose OmniOVCD, a standalone framework designed for OVCD. By leveraging the decoupled output heads of SAM 3, we propose a Synergistic Fusion to Instance Decoupling (SFID) strategy. SFID first fuses the semantic, instance, and presence outputs of SAM 3 to construct land-cover masks, and then decomposes them into individual instance masks for change comparison. This design preserves high accuracy in category recognition and maintains instance-level consistency across images. As a result, the model can generate accurate change masks. Experiments on four public benchmarks (LEVIR-CD, WHU-CD, S2Looking, and SECOND) demonstrate SOTA performance, achieving IoU scores of 67.2, 66.5, 24.5, and 27.1 (class-average), respectively, surpassing all previous methods.

Figures (5)

• Figure 1: Architecture comparison.
• Figure 2: Compared with existing methods, OmniOVCD achieves superior IoU scores across multiple category benchmarks.
• Figure 3: (a) shows the overview of OmniOVCD framework. The model takes bi-temporal images and corresponding text prompts to generate initial masks via SAM 3. These masks are then separated into instance masks for instance-level comparison, which produce the accurate change detection mask. (b) shows the multi-head outputs fusion strategy. This strategy fuses the semantic and instance head outputs from SAM 3 and uses the presence head outputs for filtering. This approach effectively improves the accuracy of single-image segmentation.
• Figure 4: Comparison of GPU memory usage, inference efficiency and IoU performance on an 3090 GPU.
• Figure 5: Visual comparisons of OmniOVCD with other state-of-the-art methods for open-vocabulary change detection.