Open-Vocabulary Semantic Segmentation in Remote Sensing via Hierarchical Attention Masking and Model Composition
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914357468725248 |
|---|---|
| author | Heidarianbaei, Mohammadreza Dorozynski, Mareike Kanyamahanga, Hubert Mehltretter, Max Rottensteiner, Franz |
| author_facet | Heidarianbaei, Mohammadreza Dorozynski, Mareike Kanyamahanga, Hubert Mehltretter, Max Rottensteiner, Franz |
| contents | In this paper, we propose ReSeg-CLIP, a new training-free Open-Vocabulary Semantic Segmentation method for remote sensing data. To compensate for the problems of vision language models, such as CLIP in semantic segmentation caused by inappropriate interactions within the self-attention layers, we introduce a hierarchical scheme utilizing masks generated by SAM to constrain the interactions at multiple scales. We also present a model composition approach that averages the parameters of multiple RS-specific CLIP variants, taking advantage of a new weighting scheme that evaluates representational quality using varying text prompts. Our method achieves state-of-the-art results across three RS benchmarks without additional training. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_23869 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Open-Vocabulary Semantic Segmentation in Remote Sensing via Hierarchical Attention Masking and Model Composition Heidarianbaei, Mohammadreza Dorozynski, Mareike Kanyamahanga, Hubert Mehltretter, Max Rottensteiner, Franz Computer Vision and Pattern Recognition In this paper, we propose ReSeg-CLIP, a new training-free Open-Vocabulary Semantic Segmentation method for remote sensing data. To compensate for the problems of vision language models, such as CLIP in semantic segmentation caused by inappropriate interactions within the self-attention layers, we introduce a hierarchical scheme utilizing masks generated by SAM to constrain the interactions at multiple scales. We also present a model composition approach that averages the parameters of multiple RS-specific CLIP variants, taking advantage of a new weighting scheme that evaluates representational quality using varying text prompts. Our method achieves state-of-the-art results across three RS benchmarks without additional training. |
| title | Open-Vocabulary Semantic Segmentation in Remote Sensing via Hierarchical Attention Masking and Model Composition |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2602.23869 |