4KLSDB is a native-4K image dataset with 129,484 train / 2,000 val / 1,984 test images, spanning nature, urban scenes, people, food, artwork, CGI, animals, and architecture. It supports both image restoration (super-resolution) and 4K text-to-image generation.
Quick links · 🌐 Project page · 💻 Code (GitHub) · 📄 Paper (arXiv) · 🤗 Dataset · 🧱 Checkpoints… See the full description on the dataset page:
https://huggingface.co/datasets/SingleBicycle/4KLSDB.
