kshitijd/mmu-norm-visual
BTSbot alert triplets and Galaxy10 RGB — derived visual images (release v1) This repository contains two separate visual-image data for image-based tasks. Their values do not include calibrated photometry. BTSbot (17,394 alert rows for 8,697 matched ZTF objects) Each row contains a 63×63 science/reference/difference triplet. The upstream processing normalized these images and converted NaNs and padding to zero. This L1 dataset preserves that representation… See the full description on the dataset page: https://huggingface.co/datasets/kshitijd/mmu-norm-visual.
BTSbot alert triplets and Galaxy10 RGB — derived visual images (release v1)
This repository contains two separate visual-image data for image-based tasks. Their values do not include calibrated photometry.
BTSbot (17,394 alert rows for 8,697 matched ZTF objects)
Each row contains a 63×63 science/reference/difference triplet. The upstream processing normalized these images and converted NaNs and padding to zero. This L1 dataset preserves that representation unchanged. There is one row per alert, about two per object on average, with alert metadata retained. Causal and prefix-prediction views should exclude future-derived attributes such as peak measurements and final class.
Galaxy10 / gz10 (7,096 matched galaxies)
Each row contains a 256×256 RGB visualization image as uint8 PNG bytes and a hard class label. Physical fluxes and uncertainties are unavailable, and the labels use the Galaxy10 hard-vote ontology.
Credit: BTSbot (Rehemtulla et al. 2024) / ZTF; Galaxy10 (astroNN) / Galaxy Zoo + DECaLS. Sourced via Multimodal Universe.