ODI-Möbius (ODIM)
收藏arXiv2025-09-30 收录
下载链接：
http://vlislab22.github.io/OmniZoomer/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为ODI-Möbius，旨在为OmniZoomer模型训练提供支持，并以监督方式比较不同方法。该数据集基于ODI-SR数据集构建，包含了在训练过程中应用的各种Möbius变换的图像。在验证和测试阶段，使用了固定的变换矩阵。数据集规模包括1191张训练图像，100张验证图像以及100张测试图像。其任务是监督学习，旨在使用Möbius变换来转换全向图像。
The dataset, named ODI-Möbius, is developed to support the training of the OmniZoomer model and facilitate supervised comparison of various methods. It is constructed based on the ODI-SR dataset, and contains images that have undergone various Möbius transformations during the training phase. Fixed transformation matrices are employed during both the validation and test phases. The dataset consists of 1,191 training images, 100 validation images, and 100 test images. Its core task is supervised learning, which aims to transform omnidirectional images using Möbius transformations.
