AI

LandslideFM: A Foundation Model Towards Unified Landslide Perception and Understanding

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 11/11/2026, 05:03:23 UTC.

Briefings

LandslideFM is a unified foundation model designed to integrate visual perception with contextual interpretation for landslide analysis. It employs a Source-Agreement Graduated Trust semi-supervised learning framework to handle limited labeled data while utilizing abundant unlabeled information. The model includes LandslideVL, a benchmark combining visual imagery with textual prompts, and achieves superior performance compared to general-purpose Vision-Language Models in specialized landslide reasoning. The research was published on arXiv in November 2026.

Key points

• LandslideFM uses a Source-Agreement Graduated Trust semi-supervised learning framework to adapt to multiple visual tasks using limited labeled data and abundant unlabeled data.
• LandslideFM surpasses general-purpose Vision-Language Models in specialized landslide reasoning and scene analysis, supported by experiments on public datasets.

Context

This research addresses the limitations of existing landslide detection models, which often focus on isolated visual tasks and lack deep analytical capabilities. LandslideFM represents the first unified foundation model connecting visual perception with contextual interpretation, offering explicit spatial guidance to Vision-Language Models for comprehensive landslide reasoning.

Limitations

The study relies on public datasets and limited labeled data, with no additional background or context provided. Results are based on experiments conducted on LandslideVL, and specific performance metrics are not detailed in the evidence.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 11/11/2026, 05:06:41 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeid72vkxts4wzbglitmhvjyhfwuqkpa5egsgb6g2d4hf7xtwacfhyq