TAMP-Street
收藏arXiv2025-09-30 收录
下载链接：
https://drive.google.com/drive/folders/1frK37q4CZ2tDUc5_x-65N2wOXEd2oJcL?usp=sharing
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为TAMP-Street，旨在评估在时间变化图像修复（TAMP）设置下的图像修复模型性能，它使用了从同一场景在不同时间显著拍摄的街道视图图像对。该数据集基于VL-CMU-CD数据集构建，包含总共1362对尺寸为1024×768的图像对，并包括用于测试的具有随机像素损坏的遮罩对。规模上，该数据集包含1362对图像，其任务定位于时间变化图像修复。
The dataset named TAMP-Street is designed to assess the performance of image inpainting models under the temporal-variant image inpainting (TAMP) setting. It employs pairs of street-view images of the same scene captured at significantly different times. This dataset is constructed based on the VL-CMU-CD dataset, including a total of 1,362 image pairs with a resolution of 1024×768, and also incorporates mask pairs with random pixel corruption for testing purposes. In terms of scale, the dataset contains 1,362 image pairs, with its task targeted at temporal-variant image inpainting.
提供机构：
Assembled based on existing image and mask datasets
