RAMP Astro benchmark: Galaxy deblending - Authors: Bertrand Rigaud - Alexandre Boucaud - - Description: The challenge uses 20 000 images (128x128 pixels, 32 bits) - Size: The dataset (2 Gb) is divided into 12 000 images for the training, 4 000 for the validation (during training) and 4 000 images for the final evaluation of the model. - File type: 4 npy files: data_train.npy, data_test.npy, labels_test.npy, labels_train.npy - License: This work is released under a BSD-3 license - Tag: - Prace dataset location: - url: - Download url: - Download script: ./download_data.py ===================================================================