SpecLess

SpecLess is an open-source Python framework for photometry-based machine-learning classification of galaxies.

The project is designed to provide reusable pretrained models, photometric feature-engineering tools, survey-validation workflows, and a collaborative framework for galaxy classification without requiring spectroscopy at inference time.

Current status

SpecLess is currently under active development.

Model 1: Galaxy Morphology

Binary classification:

• Elliptical
• Spiral

Primary classifier:

• XGBoost

Required photometric bands:

• SDSS u, g, r, i, z
• WISE W1, W2

SpecLess automatically constructs the 21 colour features used by the pretrained classifier, producing a total of 28 photometric features.

The current Model 1 classifier was trained on 142,212 galaxies and evaluated on an independent test sample of 35,554 galaxies.

Development installation

Clone the repository and install in editable mode:

pip install -e .
import pandas as pd
from specless_astro import MorphologyClassifier

catalog = pd.read_csv("catalog.csv")

model = MorphologyClassifier()

result = model.predict(catalog)

print(
    result[
        [
            "specless_label",
            "p_elliptical",
            "p_spiral",
            "confidence",
        ]
    ]
)