mb-change_cls_hirise

A Mars image classification dataset for planetary science research.

Dataset Metadata

• License: CC-BY-4.0 (Creative Commons Attribution 4.0 International)
• Version: 1.0
• Date Published: 2025-05-22
• Cite As: TBD

Classes

This dataset contains the following classes:

• 0: no_change
• 1: change

Statistics

• train: 6206 images
• test: 1340 images
• val: 1340 images
• few_shot_train_2_shot: 4 images
• few_shot_train_1_shot: 2 images
• few_shot_train_10_shot: 20 images
• few_shot_train_5_shot: 10 images
• few_shot_train_15_shot: 30 images
• few_shot_train_20_shot: 40 images
• partition_train_0.01x_partition: 31 images
• partition_train_0.02x_partition: 62 images
• partition_train_0.50x_partition: 1565 images
• partition_train_0.20x_partition: 626 images
• partition_train_0.05x_partition: 156 images
• partition_train_0.10x_partition: 313 images
• partition_train_0.25x_partition: 782 images

Few-shot Splits

This dataset includes the following few-shot training splits:

• few_shot_train_2_shot: 4 images
• few_shot_train_1_shot: 2 images
• few_shot_train_10_shot: 20 images
• few_shot_train_5_shot: 10 images
• few_shot_train_15_shot: 30 images
• few_shot_train_20_shot: 40 images

Few-shot configurations:

• 2_shot.csv
• 1_shot.csv
• 10_shot.csv
• 5_shot.csv
• 15_shot.csv
• 20_shot.csv

Partition Splits

This dataset includes the following training data partitions:

• partition_train_0.01x_partition: 31 images
• partition_train_0.02x_partition: 62 images
• partition_train_0.50x_partition: 1565 images
• partition_train_0.20x_partition: 626 images
• partition_train_0.05x_partition: 156 images
• partition_train_0.10x_partition: 313 images
• partition_train_0.25x_partition: 782 images

Usage

from datasets import load_dataset

dataset = load_dataset("Mirali33/mb-change_cls_hirise")

Format

Each example in the dataset has the following format:

{
  'image': Image(...),  # PIL image
  'label': int,         # Class label
}