🤖 AI Summary
This work addresses the scarcity of training-free, open-vocabulary methods for semantic segmentation in remote sensing imagery, a task typically hindered by the high cost of pixel-level annotations. The authors propose DinoSplat-OV, a novel framework that, for the first time, directly leverages DINOv3 for open-vocabulary segmentation of remote sensing images without requiring fine-tuning or additional pretraining. To tackle the challenges posed by the dense, multi-scale, and large-size nature of such imagery, the method integrates text-guided denoising Laplacian propagation, RGB-guided anisotropic feature aggregation, Gaussian lattice upsampling, and a global anchor sliding-window mechanism. Evaluated on UDD5, DOTA, and LoveDA benchmarks, DinoSplat-OV achieves performance on par with or superior to existing zero-shot approaches, thereby filling a critical gap in the application of DINO-based models to this domain.
📝 Abstract
Remote sensing semantic segmentation is hindered by costly pixel-level annotations, motivating training-free open-vocabulary methods. Recently, the recent release of DINOv3 brings DINO.txt, which equips the standalone DINO backbone with image-text contrastive learning and thus opens up the possibility of open-vocabulary segmentation. We propose DinoSplat-OV, a training-free framework that adapts DINOv3 to remote sensing without fine-tuning or additional pretraining. Targeting the dense distribution, multi-scale nature, and large size of remote sensing imagery, we design two core modules. Its Text-aware Laplacian Propagation module de-noises patch-level predictions by combining textual semantic affinities with local visual similarity, improving regional consistency while preserving boundaries. Its Gaussian Splatting Upsampling module reconstructs pixel-level features through RGB-guided anisotropic aggregation and test-time optimization. A global-anchor sliding-window strategy further supports large-scale imagery. Experiments on UDD5, DOTA, and LoveDA demonstrate competitive or superior performance over existing training-free methods, effectively filling the gap of DINO-series models in training-free open-vocabulary segmentation and providing a viable new path for further advances in this direction.