Galaxy Zoo DECaLS: 8-Class Morphological Labels for 100K+ Images

Between 100,000 and 1,000,000 galaxy images with 8 morphological classification labels derived from the Galaxy Zoo DECaLS project. Created by mrJordi0 and based on the 2022 work by Walmsley et al., it is specifically formatted for training deep learning architectures like Vision Transformers. The data is provided in optimized Parquet format for efficient processing in machine learning pipelines.

Use Cases

• Training Efficient Vision Transformers (ViT) for automated galaxy morphological classification
• Benchmarking image classification algorithms on astronomical datasets using the 8 morphological labels
• Astrophysical research into galaxy population statistics using DECaLS survey imagery

Strengths

• Scale of 100,000 to 1,000,000 records
• 8 distinct morphological label categories
• Derived from the peer-reviewed Galaxy Zoo DECaLS project (Walmsley et al., 2022)
• Provided in optimized-parquet format for high-speed I/O

Limitations

• Morphological labels are restricted to 8 classes, which may simplify complex galaxy structures
• Geographic bias is inherent to the DECaLS survey footprint
• Labels are derived from crowdsourced classifications which may contain noise compared to professional astronomer consensus

Provenance

Source

Galaxy Zoo DECaLS project (Walmsley et al., 2022)

Collection Method

Derived from the Dark Energy Camera Legacy Survey (DECaLS) and annotated via the Galaxy Zoo crowdsourcing platform

Freshness

Last updated February 2026.

Geography

Celestial (DECaLS survey footprint)