MMFace-DiT-Datasets

MMFace-DiT Dataset: Multimodal Face Generation Benchmarks This repository contains the multimodal conditioning data and high-quality captions for MMFace-DiT, accepted to CVPR 2026. This dataset provides the necessary spatial (masks, sketches) and semantic (VLM-enriched captions) pairs to enable high-fidelity, controllable face synthesis. 📂 Dataset Components The dataset is organized to be plug-and-play with the MMFace-DiT repository: Celeb_Dataset/:… See the full description on the dataset page: https://huggingface.co/datasets/BharathK333/MMFace-DiT-Datasets.