1. Introduction

GVLM v1.1 is a multi-modal version of GVLM. It contains geo-referenced remote sensing images with high-quality labels.

2. Event Details

This dataset covers 7 typical landslide events in recent years. Sites are standardized as "City/Region, Country" with their dates following the YYYY-MM-DD format.

ID	Site (City/Region, Country)	Image Size (Pixels)	Trigger	Event Date
1	Mtauchira, Malawi	5376 x 6144	Cyclone	2023-03
2	Wuxie, China	10496 x 8960	Rainstorm	2021-06-10
3	Longchuan, China	9216 x 7680	Rainstorm	2019-06-13
4	Te Haroto, New Zealand	10496 x 11264	Cyclone	2023-02-21
5	São Sebastião, Brazil	10496 x 5888	Urban Growth	2023-02
6	Nippes, Haiti	11008 x 6144	Earthquake	2021-08-14
7	Enshi, China	7168 x 5376	Rainfall	2021-06-21

3. Technical Specifications

Each event provides multi-temporal imagery: Pre-event (img1) and Post-event (img2).

• High-Resolution Optical (HR-RGB)

• Resolution: ~1.2 m/pixel

• Bands: 3 (Red, Green, Blue) | uint8

• Multi-Spectral (MS)

• Resolution: ~10 m/pixel

• Bands: 4 (B2: Blue, B3: Green, B4: Red, B8: NIR) | uint8

• Synthetic Aperture Radar (SAR)

• Resolution: ~10 m/pixel

• Polarization: Dual-pol (VV, VH) | int8

4. Directory Structure

GVLM_1.1_8bit/
├── [Site_Name]/
    ├── img1_HR.tif   # Pre-event High-Resolution RGB
    ├── img1_MS.tif   # Pre-event Multi-Spectral
    ├── img1_SAR.tif  # Pre-event SAR
    ├── img2_HR.tif   # Post-event High-Resolution RGB
    ├── img2_MS.tif   # Post-event Multi-Spectral
    ├── img2_SAR.tif  # Post-event SAR
    └── label.tif     # Landslide Ground Truth (Binary Mask)

5. Download

GVLM_1.1_8bit.zip from Baidu Netdisk or Google Drive:

• Baidu Netdisk: Access Link (PWD: 7u8u)
• Google Drive: Access Link

6. Acknowledgement

We would like to thank the Google Earth platform and ESA Sentinel missions for providing the remote sensing images.