Masked Autoencoder for Euclid Images

This masked autoencoder (MAE) is trained to reconstruct Euclid galaxy images where 90% of the image is masked. The trained model shows superhuman performance at reconstruction.

UPDATED VERSION AVAILABLE

This version is trained on Euclid RR2 (3M images). We later trained an otherwise-identical model on Euclid DR1 (13.6M images). You should use that instead - it's strictly better!

Go to mwalmsley/euclid-dr1-mae for full details and example code.

We'll leave this model up for reproducibility, but seriously, please use the DR1 version!