paper-with-me

Dark Zurich

홈페이지 · 논문 57편

Dark Zurich is an image dataset containing a total of 8779 images captured at nighttime, twilight, and daytime, along with the respective GPS coordinates of the camera for each image. These GPS annotations are used to construct cross-time-of-day correspondences, i.e., to match each nighttime or twilight image to its daytime counterpart. These attributes allow the usage of Dark Zurich as a dataset to build models and systems that perform: 1) domain adaptation (unsupervised, weakly supervised or semi-supervised), e.g. for semantic segmentation or object detection, 2) image translation / style transfer to different times of day, 3) robust image matching / visual localization across diverse domains, and 4) other visual perception tasks that are central for autonomous vehicles and other robotic applications.

Images

벤치마크

Semantic Segmentation on Dark Zurich 결과 28개
Unsupervised Semantic Segmentation on Dark Zurich 결과 3개
Source-Free Domain Adaptation on Cityscapes to Dark Zurich 결과 2개