mespinosami/copgen-vaes
Updated
"COP-GEN: Latent Diffusion Transformer for Copernicus Earth Observation Data – Generation Stochastic by Design" (Model checkpoints, assets, datasets)
Note [MODEL] COP-GEN Variational Autoencoders - DEM (64x64) - LULC (192x192) - S1RTC_VV_VH (192x192) - S2L1C_B4_3_2_8 (192x192) - S2L1C_cloud_mask (192x192) - S2L1C_B1_9_10 (32x32) - S2L1C_B5_6_7_8A_11_12 (96x96) - S2L2A_B4_3_2_8 (192x192) - S2L2A_B1_9 (32x32) - S2L2A_B5_6_7_8A_11_12 (96x96)
Note [MODEL] COP-GEN Base model, 500k steps
Note [DATASET] Small subset of images for rapid development (Edinburgh Region)
Note [DATASET] Test images (495 images)
Note [DATASET] Hold out 1k images.
Note [DATASET]