BACK TO LOG

MISSION LOG #006

PROJECT LOGS

How AstroVision Works

Jun 18, 20267 min

JWST image pipeline

From archive to array

FITS files are downloaded from MAST, cleaned with Astropy and OpenCV, then written to HDF5 so training batches load quickly.

The model

A convolutional network classifies the cutouts. Keeping the architecture small made iteration fast during the hackathon.

Serving it

A Flask interface lets a user upload an image and see the predicted class with the processed preview beside it.

RELATED ENTRIES