AstroDash is a Django web application for ML-based supernova spectrum classification using deep learning. It provides both a web interface and a REST API for classifying astronomical spectra using DASH CNN, Transformer, and user-uploaded models.
- Single spectrum classification — upload a spectrum file or reference a supernova by name from the Open Supernova Catalog
- Batch processing — classify multiple spectra via ZIP file upload
- Multiple classifier models — DASH CNN, Transformer, and user-uploaded TorchScript models
- Spectral twins explorer — find similar spectra using embeddings and UMAP
- Redshift estimation — estimate redshift using DASH templates
- REST API — full API for programmatic access
AstroDash follows a layered architecture:
- Web layer — Django views and templates with Bootstrap for the interactive UI
- API layer — Django REST Framework endpoints under
/astrodash/api/v1/ - Domain services — business logic for spectrum processing, classification, and model management
- ML infrastructure — PyTorch-based classifiers, preprocessing, and template handling
- Async processing — Celery workers with Redis for batch classification tasks
The application is containerized with Docker and deployed to Kubernetes on Jetstream2 using ArgoCD for GitOps.
| Environment | URL |
|---|---|
| Production | https://astrodash.scimma.org |
| Development | https://astrodash-dev.scimma.org |
- API Reference — endpoints, data formats, error handling
- User Guides — getting started, code examples
- Developer Guide — local setup, contributing, running tests
- Admin Guide — managing data files and deployments
- Contributing a Classifier — how to add a new ML model
See CONTRIBUTING.md for guidelines and the Developer Guide for local setup instructions.
This project is supported by National Science Foundation grants OAC-1841625, OAC-1934752, OAC-2311355, AST-2432428.
AstroDash builds on the DASH spectral classification tool and the original Blast web application.
See acknowledgements for full contributor, data source, and software credits.