A Benchmark Study of Segmentation Models and Adaptation Strategies for Landslide Detection from Satellite Imagery

2026-04-17

AI Summary

This study addresses the lack of systematic evaluation of modern segmentation models and their fine-tuning strategies for landslide detection. Under a unified experimental protocol, we comprehensively assess the performance of convolutional neural networks, Transformer-based architectures, and large-scale pretrained foundation models on high-resolution satellite imagery for landslide detection, while comparing full fine-tuning against parameter-efficient approaches such as LoRA and AdaLoRA. Our results demonstrate that Transformer models significantly outperform conventional CNNs, and that parameter-efficient fine-tuning achieves comparable accuracy to full fine-tuning while reducing trainable parameters by up to 95%, alongside strong out-of-distribution generalization. This work is the first to systematically reveal the practical value of efficient fine-tuning in remote sensing–based disaster detection.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Landslide detection from high resolution satellite imagery is a critical task for disaster response and risk assessment, yet the relative effectiveness of modern segmentation architectures and finetuning strategies for this problem remains insufficiently understood. In this work, we present a systematic benchmarking study of convolutional neural networks, transformer based segmentation models, and large pre-trained foundation models for landslide detection. Using the Globally Distributed Coseismic Landslide Dataset (GDCLD) dataset, we evaluate representative CNN- and transformer-based segmentation models alongside large pretrained foundation models under consistent training and evaluation protocols. In addition, we compare full fine-tuning with parameter-efficient fine-tuning methods, including LoRA and AdaLoRA, to assess their performance efficiency tradeoffs. Experimental results show that transformer-based models achieve strong segmentation performance, while parameter efficient finetuning reduces trainable parameters by up to 95% with comparable accuracy to full finetuning. We further analyze generalization under distribution shift by comparing validation and held-out test performance.

Problem

• landslide detection
• satellite imagery
• segmentation models
• fine-tuning strategies
• benchmarking

Innovation

• transformer-based segmentation
• parameter-efficient fine-tuning
• LoRA
• landslide detection
• foundation models