# Mars-Bench Atmospheric Dust Classification, HiRISE EDR Tiles (Dusty vs Not Dusty)

## Facts

- Provider: Mirali Purohit, Hannah Kerner et al. (Mars-Bench, Arizona State University)
- License: CC-BY-4.0 (https://arxiv.org/html/2510.24010v1)
- Size: 20 files
- Format: CSV, Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/Mirali33/mb-atmospheric_dust_cls_edr
- Access: public, readable without an account
- DOI: https://doi.org/10.48550/arXiv.2510.24010
- Category: Astronomy & Space
- Source: https://huggingface.co/datasets/Mirali33/mb-atmospheric_dust_cls_edr
- Page: https://scigantic.com/archive/2b4fd06e-e566-593d-80c2-89bb0cfd338d
- Keywords: Astronomy & Space, CSV, Parquet, Hugging Face

## About

Binary image classification of Martian atmospheric dust: 100 x 100 pixel tiles cut from MRO HiRISE Experiment Data Record (EDR) images, the raw, uncalibrated and unstitched instrument frames, each labeled dusty or not_dusty. Built from the Doran (2019) HiRISE dust dataset and class-balanced, with 9,817 training, 4,969 validation and 5,214 test tiles, plus 1- to 20-shot training subsets for few-shot evaluation. A sibling set (mb-atmospheric_dust_cls_rdr) holds the same task on Reduced Data Record tiles, so the two can be compared for the effect of calibration. Stored as Parquet shards under data/ (image plus integer label), with annotation.csv and mapping.json; load with datasets.load_dataset("Mirali33/mb-atmospheric_dust_cls_edr"). About 150 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 Doran (2019).

## Load the data

```python
import pandas as pd
df = pd.read_csv("hf://datasets/Mirali33/mb-atmospheric_dust_cls_edr/annotation.csv")
df.head()
```

## Data structure

File types in sampled files: CSV, Parquet

Example file: `annotation.csv`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/2b4fd06e-e566-593d-80c2-89bb0cfd338d
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/2b4fd06e-e566-593d-80c2-89bb0cfd338d
- Over MCP: `get_schema_card` with id `2b4fd06e-e566-593d-80c2-89bb0cfd338d` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [CSV datasets](https://scigantic.com/topics/csv)
- [Astronomy & Space datasets](https://scigantic.com/topics/astronomy-and-space)
