Vision Foundation Models (VFMs), while powerful, often struggle in Remote Sensing (RS) segmentation tasks when combined with existing Parameter-Efficient Fine-Tuning (PEFT) methods. We observe that this limitation primarily arises from their inability to effectively handle the pervasive artifacts in RS imagery. To address this, we introduce Earth-Adapter, the first PEFT method specifically designed for RS artifact mitigation. Earth-Adapter introduces a novel Frequency-Guided Mixture of Adapters (MoA) approach, structured around a ''divide and conquer" strategy. It first utilizes Discrete Fourier Transformation (DFT) to "divide" features into distinct frequency components, thereby effectively isolating artifact-related information from semantic signals. Subsequently, to ''conquer" these artifact, MoA independently optimizes features within different subspaces and dynamically assigns weights via a router to aggregate the refined representations. This enables adaptive refinement of the VFM’s representation space to mitigate the impact of artifacts. This simple yet highly effective PEFT method demonstrably mitigates artifacts and significantly enhances VFMs performance on RS segmentation tasks. Extensive experiments demonstrate Earth-Adapter's effectiveness on in-domain semantic segmentation (SS), as well as Domain Adaptive (DA) and Domain Generalized (DG) semantic segmentation tasks. Compared with the baseline Rein, Earth-Adapter significantly improves mIoU by 1.2% in SS, 9.0% in DA, and 3.1% in DG benchmarks. Our code and weights will be released soon.

Due to the substantial domain gaps in Remote Sensing (RS) images that are characterized by variabilities such as location, wavelength, and sensor type, Remote Sensing Domain Generalization (RSDG) has emerged as a critical and valuable research frontier, focusing on developing models that generalize effectively across diverse scenarios. However, research in this area remains underexplored: (1) Current cross-domain methods primarily focus on Domain Adaptation (DA), which adapts models to predefined domains rather than to unseen ones; (2) Few studies target the RSDG issue, especially for semantic segmentation tasks. Existing related models are developed for specific unknown domains, struggling with issues of underfitting on other unseen scenarios; (3) Existing RS foundation models tend to prioritize in-domain performance over cross-domain generalization. To this end, we introduce the first vision foundation model for RSDG semantic segmentation, CrossEarth. CrossEarth demonstrates strong cross-domain generalization through a specially designed data-level Earth-Style Injection pipeline and a model-level Multi-Task Training pipeline. In addition, for the semantic segmentation task, we have curated an RSDG benchmark comprising 32 semantic segmentation scenarios across various regions, spectral bands, platforms, and climates, providing comprehensive evaluations of the generalizability of future RSDG models. Extensive experiments on this collection demonstrate the superiority of CrossEarth over existing state-of-the-art methods.

Contrastive Language-Image Pre-training (CLIP) has achieved success on multiple downstream tasks by aligning image and text modalities. However, the nature of global contrastive learning limits CLIP's ability to comprehend compositional concepts, such as relations and attributes. Although recent studies employ global hard negative samples to improve compositional understanding, these methods significantly compromise the model's inherent general capabilities by forcibly distancing textual negative samples from images in the embedding space. To overcome this limitation, we introduce a Decoupled Global-Local Alignment (DeGLA) framework that improves compositional understanding while substantially mitigating losses in general capabilities. To optimize the retention of the model's inherent capabilities, we incorporate a self-distillation mechanism within the global alignment process, aligning the learnable image-text encoder with a frozen teacher model derived from an exponential moving average. Under the constraint of self-distillation, it effectively mitigates the catastrophic forgetting of pretrained knowledge during fine-tuning. To improve compositional understanding, we first leverage the in-context learning capability of Large Language Models (LLMs) to construct about 2M high-quality negative captions across five types. Subsequently, we propose the Image-Grounded Contrast (IGC) loss and Text-Grounded Contrast (TGC) loss to enhance vision-language compositionally. Experimental results across both general and compositional reasoning tasks validate the effectiveness of the DeGLA framework. Our code is released at https://github.com/xiaoxing2001/DeGLA.

Semantic segmentation of remote sensing images aims to achieve pixel-level semantic category assignment for input images. This task has achieved significant advances with the rapid development of deep neural network. Most current methods mainly focus on effectively fusing the low-level spatial details and high-level semantic cues. Other methods also propose to incorporate the boundary guidance to obtain boundary preserving segmentation. However, current methods treat the multi-level feature fusion and the boundary guidance as two separate tasks, resulting in sub-optimal solutions. Moreover, due to the large inter-class difference and small intra-class consistency within remote sensing images, current methods often fail to accurately aggregate the long-range contextual cues. These critical issues make current methods fail to achieve satisfactory segmentation predictions, which severely hinder downstream applications. To this end, we first propose a novel boundary guided multi-level feature fusion module to seamlessly incorporate the boundary guidance into the multi-level feature fusion operations. Meanwhile, in order to further enforce the boundary guidance effectively, we employ a geometric-similarity-based boundary loss function. In this way, under the explicit guidance of boundary constraint, the multi-level features are effectively combined. In addition, a channel-wise correlation guided spatial-semantic context aggregation module is presented to effectively aggregate the contextual cues. In this way, subtle but meaningful contextual cues about pixel-wise spatial context and channel-wise semantic correlation are effectively aggregated, leading to spatial-semantic context aggregation. Extensive qualitative and quantitative experimental results on ISPRS Vaihingen and GaoFen-2 datasets demonstrate the effectiveness of the proposed method.