MIMIC-1M
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/RAIVNLab/MIMIC
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了130万对多视角图像，这些图像来源于现实世界视频和合成环境。该数据集特别设计用来整合多视角图像，无需额外的标注，从而能够有效地训练模型，以应对密集几何任务。规模达到了130万对图像，其任务涵盖了遮罩图像建模和表征学习。
This dataset contains 1.3 million pairs of multi-view images sourced from real-world videos and synthetic environments. It is specifically designed to process multi-view visual data without requiring additional annotations, enabling efficient training of models for dense geometric tasks. With such a scale, its supported tasks include masked image modeling and representation learning.
提供机构：
RAIVN Lab
