Dataset Catalog

This dataset is part of T2I-CompBench and contains text files for multiple configurations and splits, primarily for textual data. The dataset includes various configurations such as '3d_spatial_train', 'color_val', 'complex_train', 'shape_val', 'texture_train', each with specific features and splits. Features are mainly text data, and splits contain training and validation sets of varying sizes. The dataset serves as a comprehensive benchmark for text‑to‑image generation and is licensed under the MIT License. It is obtained from a GitHub repository and uploaded to the Hugging Face Hub.