WILD-SAM is a fine-tuned SAM variant using phase-aware MoE adapters and wavelet subband enhancement that achieves state-of-the-art landslide detection on wrapped InSAR data.
Rsprompter: Learning to prompt for remote sensing instance seg- mentation based on visual foundation model,
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CV 2verdicts
UNVERDICTED 2representative citing papers
Combining semantic and geometric prompts with light fine-tuning gives the best SAM3 performance on remote sensing segmentation, while text-only prompting lags especially on irregular shapes.
citing papers explorer
-
WILD-SAM: Phase-Aware Expert Adaptation of SAM for Landslide Detection in Wrapped InSAR Interferograms
WILD-SAM is a fine-tuned SAM variant using phase-aware MoE adapters and wavelet subband enhancement that achieves state-of-the-art landslide detection on wrapped InSAR data.
-
On the Effectiveness of Textual Prompting with Lightweight Fine-Tuning for SAM3 Remote Sensing Segmentation
Combining semantic and geometric prompts with light fine-tuning gives the best SAM3 performance on remote sensing segmentation, while text-only prompting lags especially on irregular shapes.