A Training-Free Vision Foundation Model Synergy Framework for Building Extraction in High-Resolution Remote Sensing Images

Research on Intelligent Building Extraction Models Based on High-Resolution Remote Sensing Imagery

MsRE: Toward Efficient Remote Sensing Segmentation via Vision Foundation Models

Vision foundation models (VFMs) pretrained on large-scale datasets have significantly improved performance in remote sensing semantic segmentation. However, existing methods typically rely on full fine-tuning, which requires updating all model parameters. Instead of updating the full parameter set, parameter-efficient fine-tuning (PEFT) achieves competitive performance by optimizing only a small subset of parameters. Despite its success, most existing PEFT methods are mainly designed for natural image tasks and fail to account for the unique multiscale characteristics of remote sensing images. To address these challenges, we propose multi-scale cognitive feature refinement (MsRE) tuning, a novel PEFT method tailored for remote sensing semantic segmentation. In particular, MsRE captures multiscale contextual information by applying cognitive operations with different cognitive fields to intermediate features of the backbone. It then introduces a set of learnable tokens to establish interactions with features at different scales, enabling precise feature refinement and progressive feature propagation across network layers. This mechanism enhances the model’s ability to understand complex remote sensing scenes and improves downstream segmentation performance. With significantly fewer trainable parameters, MsRE provides an efficient yet effective solution for adapting VFMs to remote sensing segmentation tasks. Extensive experiments demonstrate that MsRE achieves competitive segmentation performance with substantially fewer trainable backbone parameters, providing a favorable balance between accuracy and parameter efficiency. The project is available at http://woldier.top/MsRE

Leveraging Pretrained Priors for Weakly Supervised Semantic Segmentation of Remote Sensing Images

Abstract. Semantic segmentation of remote sensing imagery (RSI) is essential for urban mapping, land-use monitoring, and many other domains. However, pixel-level annotation is expensive, making weakly supervised semantic segmentation (WSSS) that relies on image-level labels an attractive alternative. Pre-trained models provide strong priors from large-scale learned representations, making them beneficial for WSSS. However, when kept frozen, they often produce sparse and misaligned class activation maps (CAMs) due to domain gaps and static inference. We propose a lightweight and efficient framework that integrates CLIP and DINO foundation models to address three challenges: (i) semantic misalignment between generic text prompts and RSI-specific visuals; (ii) static CAM quality; and (iii) incomplete object coverage. Our design includes: (1) a Textual Prototype-Aware Enrichment (TPE) module that builds an RS-specific knowledge base using large language model (LLM)-generated descriptions to enrich text prompts; (2) a Unified Semantic Relation Mining (USR) module that fuses learnable adapter features with CLIP attention and DINO affinity for online CAM refinement; and (3) a Visual Prototype-Aware Enrichment (VPE) module, which maintains momentum visual prototypes to complete regions and sharpen boundaries. By freezing the CLIP and DINO backbones and optimizing only lightweight adapter and decoder modules, the proposed framework reduces the number of trainable parameters while achieving competitive performance. Experimental on iSAID and ISPRS Potsdam datasets demonstrate the effectiveness of the proposed framework, achieving 38.01% mIoU on iSAID dataset and 47.01% mIoU with 66.89% overall accuracy on Potsdam dataset.

An Adaptive Network for Enhanced Vision and Semantic Cooperative Understanding in Aerial Object Detection

Object detection in remote sensing imagery faces challenges such as extreme scale variations and complex backgrounds. Although current methods have made significant strides in visual feature extraction, their predominant focus remains on the image itself, overlooking the potential of integrating external knowledge. To address this limitation, we introduce the knowledge-aware network with region-adaptive fusion for detection (KARFDet), which seamlessly integrates region-specific semantic information with visual features. First, a multiscale fused kernel attention (MSFKA) module is introduced, leveraging a parallel multibranch architecture to enhance contextual feature extraction. Second, a knowledge graph semantic extraction (KGSE) module is designed, employing the random walk with restart (RWR) algorithm to transform discrete knowledge into computable semantic associations. Finally, a novel triple-order knowledge integration (TOKI) mechanism is proposed, which adaptively fuses original, second-order, and probabilistic semantic knowledge, dynamically allocating knowledge weights based on target scale characteristics. Experiments on the DIOR, NWPU VHR-10, and SIMD datasets show that KARFDet achieves mAP50 scores of 65.6%, 91.9%, and 75.8%, respectively, significantly outperforming the baseline model and establishing a new paradigm for semantic-aware detection in complex scenarios. The code is available at https://github.com/ChengXCode/KARFDet

CGSNet: Category Prior-Guided Self-Supervised Semantic Segmentation for Remote Sensing Images

Multimodal fusion methods have shown great potential in remote sensing image analysis, but existing approaches rely heavily on massive amounts of annotated data. This is not only costly and time-consuming but also prone to subjective bias. To address this issue, we propose a category-prior-based self-supervised framework, CGSNet, which uses category prior maps extracted from multispectral images as supervisory signals for end-to-end training. An adaptive confidence-weighted pseudo-label generation mechanism is designed to alleviate noise and errors in prior maps by replacing binary labels with continuous confidence maps, enabling the learning of uncertain interclass features. In addition, a multispectral feature-guided refinement strategy utilizes color and texture information to calibrate class transition regions and enhance the discriminative power of pseudo-labels in complex scenes. A dynamic mask selection strategy further enhances the model’s robustness and generalization capabilities through progressive learning. Experiments demonstrate that CGSNet achieves state-of-the-art performance without the need for human annotation, achieving an Mean Intersection over Union (mIoU) score of 78.46% on the Gaofen image dataset (GID) (vegetation) dataset and 79.58% on the Zurich (vegetation) dataset—12.22% and 15.07% higher than existing methods, respectively—while exhibiting strong cross-dataset zero-shot generalization capabilities. The code will be available at https://github.com/NUAALISILab

ASAR-Net: A Novel Adaptive Scale-Aware Road Extraction Network for High-Resolution Remote Sensing Images

Accurate road extraction from high-resolution remote sensing imagery plays a vital role in numerous geospatial applications, including urban planning, disaster emergency response, intelligent transportation, and map updating. However, significant variations in road width, geometry, and orientation, together with complex backgrounds such as shadows, vegetation, and occlusions, often lead to incomplete extraction and poor structural continuity. To address these challenges, this paper proposes an adaptive scale-aware road extraction network, termed ASAR-Net, which jointly improves multi-scale feature representation and structural continuity. Specifically, an Adaptive Bidirectional Enhancement Module (ABEM) is introduced in the encoder to improve the representation of roads with diverse spatial scales through adaptive scale-aware convolution and bidirectional attention. Furthermore, a Directional Fusion Module (DFM) is incorporated into the decoder to guide feature reconstruction along road orientations using dynamic snake convolution, thereby facilitating the recovery of continuous and complete road structures. Extensive experiments on two public benchmark datasets, Massachusetts Roads and DeepGlobe, demonstrate that ASAR-Net consistently outperforms several representative state-of-the-art road extraction methods in terms of mIoU and F1-score. The proposed network effectively improves both the semantic completeness and structural continuity of extracted road networks, demonstrating its robustness and effectiveness for road extraction in complex high-resolution remote sensing scenarios.