Mars-Analog-Dunes-10 Dataset

Dataset Description

Mars-Analog-Dunes-10 is a curated remote sensing dataset designed for Earth-Mars Comparative Planetology. It contains high-quality satellite imagery of Earth's sand dunes that serve as morphological analogs to features found on Mars.

Key Features

• Mars Analog Focus: Selected from Earth regions (e.g., deserts in China, Africa) known for their similarity to Martian geomorphology.
• 10 Fine-Grained Classes: Consolidates complex dune types into a streamlined 10-class system suitable for rover navigation and orbital mapping training.

Categories (10 Classes)

ID	Class Name	Description	Mars Significance
0	Sand Sheets	Extensive, flat surfaces of coarse sand with no slipfaces.	Represents areas with limited sediment supply.
1	Dome Dunes	Circular mounds lacking slipfaces.	Represents early-stage formation or high wind energy.
2	Crescentic Dunes	C-shaped ridges with slipfaces (e.g., Barchan).	Most common active dunes on Mars; wind direction indicators.
3	Linear Dunes	Elongated, straight/sinuous ridges.	Indicates complex/bidirectional wind regimes.
4	Star Dunes	Pyramidal mounds with 3+ arms.	Found in craters with chaotic wind patterns.
5	Parabolic Dunes	U-shaped dunes with arms pointing upwind.	Interpreted as moisture/cementation control on Mars.
6	Network Dunes	Interconnected, mesh-like ridges.	Reflects seasonal wind variability.
7	Obstacle Dunes	Accumulations around physical barriers.	Critical for understanding dune-crater interactions.
8	Nebkha	Small mounds anchored by shrubs/obstacles.	Earth-specific noise that must be distinguished from geologic features.
9	Others	Non-dune surfaces (Bedrock, Gravel, Agriculture).	Negative samples. Crucial for distinguishing dunes from rocky terrain (TARs).

Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("Keiyoo/Mars-Analog-Dunes-10")

# Display a sample
print(dataset['train'][0])