RSW233/

MVCap-4M

Omniview-Tuning: Boosting Viewpoint Invariance of Vision-Language Pre-training Models

Shouwei Ruan, 
Yinpeng Dong, 
Hanqing Liu, Yao Huang, 
Hang Su and 
Xingxing Wei.

This repo releases the MVCap-4M dataset introduced in our paper: "Omniview-Tuning: Boosting Viewpoint Invariance of Vision-Language Pre-training Models" (ECCV2024) Multi-View Caption (MVCap-4M) is a large-scale dataset tailored for viewpoint invariance… See the full description on the dataset page: https://huggingface.co/datasets/RSW233/MVCap-4M.