MMLSv2: A Multimodal Dataset for Martian Landslide Detection in Remote Sensing Imagery

2026-02-08

AI Summary

This study addresses the significant challenges in segmenting and detecting landslides on the Martian surface, particularly in fragmented, elongated, and small-scale scenarios. To this end, the authors introduce MMLSv2, the first multimodal remote sensing dataset specifically designed for Martian landslides, integrating seven spectral and geospatial bands—including RGB, digital elevation models, slope, and thermal inertia. Crucially, the dataset incorporates a novel geographically isolated test set to rigorously evaluate model generalization to out-of-distribution regions. Experiments demonstrate that mainstream semantic segmentation models can be stably trained on MMLSv2 and achieve competitive performance on standard splits; however, their accuracy markedly declines on the isolated test set, underscoring the benchmark’s effectiveness and difficulty in assessing spatial robustness and generalization capabilities of landslide detection models.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We present MMLSv2, a dataset for landslide segmentation on Martian surfaces. MMLSv2 consists of multimodal imagery with seven bands: RGB, digital elevation model, slope, thermal inertia, and grayscale channels. MMLSv2 comprises 664 images distributed across training, validation, and test splits. In addition, an isolated test set of 276 images from a geographically disjoint region from the base dataset is released to evaluate spatial generalization. Experiments conducted with multiple segmentation models show that the dataset supports stable training and achieves competitive performance, while still posing challenges in fragmented, elongated, and small-scale landslide regions. Evaluation on the isolated test set leads to a noticeable performance drop, indicating increased difficulty and highlighting its value for assessing model robustness and generalization beyond standard in-distribution settings. Dataset will be available at: https://github.com/MAIN-Lab/MMLS_v2

Problem

• Martian landslide detection
• multimodal remote sensing
• image segmentation
• spatial generalization
• small-scale landslides

Innovation

• multimodal dataset
• Martian landslide detection
• spatial generalization
• remote sensing imagery
• segmentation benchmark