GPIC: A Giant Permissive Image Corpus for Visual Generation
Keshigeyan Chandrasegaran, Kyle Sargent, Suchir Agarwal, Michael Jang, Michael Poli, Juan Carlos Niebles, Justin Johnson, Jiajun Wu, Li Fei-Fei
Read on arXiv →Key claim
GPIC is a massive, permissively licensed image dataset.
The paper presents GPIC, a massive dataset of 28 trillion pixels designed for visual generative modeling. It includes a diverse set of images and a benchmarking protocol, making it a valuable resource for researchers and practitioners in the field. The dataset is permissively licensed, allowing for both research and commercial use.
In plain English
The paper presents GPIC, a massive dataset of 28 trillion pixels designed for visual generative modeling. It includes a diverse set of images and a benchmarking protocol, making it a valuable resource for researchers and practitioners in the field. The dataset is permissively licensed, allowing for both research and commercial use.
The introduction of a large, permissively licensed dataset for visual generative modeling significantly extends the available resources in the field.
The paper provides a comprehensive dataset with a clear benchmarking protocol and is hosted on a reputable platform, ensuring solid reliability.