Mars-Bench Boulder Segmentation, MRO HiRISE (500 x 500 tiles)

Provider
Mirali Purohit, Hannah Kerner et al. (Mars-Bench, Arizona State University)
License
CC-BY-4.0 (source)
Size
10 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/Mirali33/mb-boulder_seg)
Access
Free and public. No account with the data provider is needed.
DOI
10.48550/arXiv.2510.24010
Category
Astronomy & Space
Source
https://huggingface.co/datasets/Mirali33/mb-boulder_seg

About Mars-Bench Boulder Segmentation, MRO HiRISE (500 x 500 tiles)

Binary semantic segmentation of boulders in MRO HiRISE images of Mars, built from the manually outlined boulders of Prieur et al. (2023). Each 500 x 500 pixel tile comes with a mask of background and boulder pixels; boulder size-frequency distributions are used to assess landing-site hazards, ejecta and regolith development. The set is small (39 training, 6 validation, 4 test tiles) and was checked by domain experts for Mars-Bench, making it a test of fine-tuning with few labels for dense small-object segmentation. A companion detection version with bounding boxes is mb-boulder_det. Parquet under data/ holds image, mask, width, height and class labels, with mapping.json; load with datasets.load_dataset("Mirali33/mb-boulder_seg"). About 20 MB. Part of Mars-Bench (Purohit et al., NeurIPS 2025 Datasets and Benchmarks track, arXiv:2510.24010), 20 standardized Mars datasets released by the authors under CC BY 4.0; cite Mars-Bench and Prieur et al. (2023).

How to load Mars-Bench Boulder Segmentation, MRO HiRISE (500 x 500 tiles) in Python

from datasets import load_dataset
ds = load_dataset("Mirali33/mb-boulder_seg")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: data/partition_train_0.10x_partition-00000-of-00001.parquet

Analyze this dataset without downloading it

On Scigantic, Mars-Bench Boulder Segmentation, MRO HiRISE (500 x 500 tiles) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Astronomy & Space datasets