The lack of access to source imagery and limited target annotations exacerbates the problem of landslide segmentation models degrading

The gap

Evidence profile

Sourced from the abstract and stated research gap and stated challenges of the source papers, classified as general, spanning 3 journals.

Supporting evidence — 4 representative gaps

• GeoPhysAdapter: Scale-Matched Geophysical Adaptation for Cross-Domain Landslide Mapping with Vision Foundation Models (2026) · arXiv

  We propose GeoPhysAdapter, which anchors on a frozen vision foundation model, restricts terrain, material, and triggering to dense spatial guidance, regional modulation, and event-timing forcing, and applies bounded adaptation at two decision units, the pixel and the candidate landslide body, reverting exactly to the visual prediction where support is insufficient.

  generalabstractevidence 5/5
  Keywords: propose geophysadapter anchors frozen vision foundation model restricts terrain material triggering dense spatial guidance regional
• DCA-UNet for Landslide Segmentation with Deformable Convolution and Aggregated Attention (2026) · Remote Sensing · doi

  Existing methods struggle with irregular geometry, substantial scale variation, and strong background interference in landslide segmentation. Prior work has not effectively addressed the challenge of heterogeneous remote-sensing conditions. There is a need for a robust and efficient landslide segmentation method that can adapt to different datasets and imaging modalities.

  generalstated research gapevidence 5/5
  Keywords: existing methods struggle irregular geometry substantial scale variation
• Diagnosing landslide segmentation across geographic clusters with limited target annotations and no access to source imagery (2026) · Scientific Reports · doi

  The lack of access to source imagery and limited target annotations exacerbates the problem of landslide segmentation models degrading when transferred to a different geographic region. Prior work has not adequately addressed this problem, and there is a need for techniques to improve the reliability of landslide segmentation models in this context.

  generalstated research gapevidence 5/5
  Keywords: lack access source imagery limited target annotations exacerbates
• Diagnosing landslide segmentation across geographic clusters with limited target annotations and no access to source imagery (2026) · Scientific Reports · doi

  The lack of access to source imagery and limited target annotations poses a significant challenge to improving the reliability of landslide segmentation models. The problem of covariate shifts induced by terrain, sensing, and illumination also poses a challenge, as it changes image appearance and thus p(x|y). The need for techniques to adapt to new geographic regions with limited annotations and no access to source imagery is a key challenge addressed by this paper.

  generalstated challengesevidence 5/5
  Keywords: lack access source imagery limited target annotations poses

Questions about this gap