README.md

metadata

dataset_info:
  features:
    - name: object_id
      dtype: int64
    - name: RGB_image
      dtype: image
    - name: VIS_image
      dtype:
        array2_d:
          shape:
            - 224
            - 224
          dtype: float32
    - name: NISP_Y_image
      dtype:
        array2_d:
          shape:
            - 224
            - 224
          dtype: float32
    - name: NISP_J_image
      dtype:
        array2_d:
          shape:
            - 224
            - 224
          dtype: float32
    - name: NISP_H_image
      dtype:
        array2_d:
          shape:
            - 224
            - 224
          dtype: float32
    - name: SED
      sequence: float32
  splits:
    - name: train
      num_bytes: 9230689639
      num_examples: 10000
    - name: test
      num_bytes: 9231368005
      num_examples: 10000
  download_size: 15929958775
  dataset_size: 18462057644
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*

Euclid galaxies for training astroPT

Here we have ~300 000 galaxy cutouts from Euclid Q1 Data Release. The cutouts are 224x224 pixel images centred on the galaxy source.

Galaxy images are accompanied by their SEDs composed with filters: MegaCam_u, HSC_g, MegaCam_r, Pan-STARRS_i, HSC_z, DECam_g, DECam_r, DECam_i, DECam_z, I_E, Y_E, J_E, and H_E.

Useful links

Paper here: TBD

Models here: https://huggingface.co/collections/msiudek/astropt-euclid-67d061928ac0a447265ac8b8

Code here: https://github.com/smith42/astroPT