MS-COCO 2014

MS-COCO 2014 is a dataset that contains images with corresponding captions, used to evaluate text-to-image generation models based on various metrics such as image quality and alignment with textual descriptions.

Papers using MS-COCO 2014 (2)

AsyncDiff: Asynchronous Timestep Conditioning for Enhanced Text-to-Image Diffusion Inference2025ProReflow: Progressive Reflow with Decomposed Velocity2025