KAITZ Multimodal Glue Dataset (metadata.csv)This file is the multimodal “glue” connecting:
a reference image (image)
a textual prompt (text)
a motion-transfer instruction (instruction)
It is meant to be the main table used by training code, enabling consistent pairing across modalities.
File: metadata.csv
Rows: 1,000 (recommended; matches number of images)
Purpose: Single source of truth for multimodal pairing.
Column
Type… See the full description on the dataset page:
https://huggingface.co/datasets/arielgd30/METADTA.
