Open Z-Pengcheng opened 6 months ago
Nice job. I have a question, why can still achieve good reconstruction results with relative poses (estimated from monocular vos) when training MVSNET with real poses? Is MVSNet independent of pose scale? Is there any way to handle it?
Nice job. I have a question, why can still achieve good reconstruction results with relative poses (estimated from monocular vos) when training MVSNET with real poses? Is MVSNet independent of pose scale? Is there any way to handle it?