AstroVision Advanced — Galaxy Morphology Classifier
Fine-tuned Zoobot (ConvNeXT Nano) for galaxy morphology classification on the Galaxy10 DECals dataset.
Model Details
- Base model: mwalmsley/zoobot-encoder-convnext_nano
- Architecture: ConvNeXT Nano encoder + LinearHead classifier
- Training framework: PyTorch Lightning
- Training dataset: Galaxy10 DECals (17,736 images)
- Epochs: 10
- Validation accuracy: 92.9%
Classes
| Label | Class |
|---|---|
| 0 | Spiral Galaxy |
| 1 | Elliptical Galaxy |
| 2 | Edge-on Disk |
| 3 | Irregular Galaxy |
| 4 | Merger |
Results
| Class | Confidence (sample) |
|---|---|
| Spiral Galaxy | 99.89% |
| Elliptical Galaxy | 81.42% |
| Edge-on Disk | 100.00% |
| Irregular Galaxy | 99.17% |
| Merger | 99.99% |
Training Data
Galaxy10 DECals — 10 original classes remapped to 5:
| Galaxy10 Original | Mapped To |
|---|---|
| Barred Spiral, Unbarred Tight Spiral, Unbarred Loose Spiral | Spiral Galaxy |
| Round Smooth, In-between Smooth, Cigar Shaped Smooth | Elliptical Galaxy |
| Edge-on without Bulge, Edge-on with Bulge | Edge-on Disk |
| Disturbed Galaxies | Irregular Galaxy |
| Merging Galaxies | Merger |
Known Limitations
The dataset has a class imbalance — Irregular Galaxy has 1,081 samples vs 6,500 for Spiral. The model is less reliable on ambiguous irregular galaxies as a result.
Links
- Live App: astrovision-advanced.streamlit.app
- GitHub: zev-walker/Astrovision-Advanced
- Zoobot: mwalmsley/zoobot