paper-with-me

COCO-Stuff

Common Objects in COntext-stuff

홈페이지 · 논문 338편

The Common Objects in COntext-stuff (COCO-stuff) dataset is a dataset for scene understanding tasks like semantic segmentation, object detection and image captioning. It is constructed by annotating the original COCO dataset, which originally annotated things while neglecting stuff annotations. There are 164k images in COCO-stuff dataset that span over 172 categories including 80 things, 91 stuff, and 1 unlabeled class. Source: Image Colorization: A Survey and Dataset Image Source: https://github.com/nightrome/cocostuff

Images

벤치마크

Unsupervised Semantic Segmentation on COCO-Stuff-27 결과 87개
Image-to-Image Translation on COCO-Stuff Labels-to-Photos 결과 45개
Semantic Segmentation on COCO-Stuff test 결과 42개
Unsupervised Semantic Segmentation with Language-image Pre-training on COCO-Stuff-171 결과 36개
Unsupervised Semantic Segmentation on COCO-Stuff-3 결과 18개
Zero-Shot Semantic Segmentation on COCO-Stuff 결과 15개
Unsupervised Semantic Segmentation on COCO-Stuff-171 결과 12개
Unsupervised Semantic Segmentation on COCO-Stuff-81 결과 12개
Unsupervised Semantic Segmentation with Language-image Pre-training on COCO-Stuff-27 결과 12개
Unsupervised Semantic Segmentation on COCO-Stuff-15 결과 9개
Open Vocabulary Semantic Segmentation on COCO-Stuff-171 결과 7개
Layout-to-Image Generation on COCO-Stuff 128x128 결과 5개
Layout-to-Image Generation on COCO-Stuff 256x256 결과 5개
Layout-to-Image Generation on COCO-Stuff 64x64 결과 5개
Real-Time Semantic Segmentation on COCO-Stuff 결과 4개
Semantic Segmentation on COCO-Stuff 결과 3개
Sketch-to-Image Translation on COCO-Stuff 결과 3개
Semantic Segmentation on COCO-Stuff full 결과 2개
Semantic Segmentation on COCO-Stuff-27 결과 2개