Open yuzhongruicn opened 4 months ago
Hi, Obtaining raw data from new videos could be done by combining different open source video processing tools, here are some suggestions but I would recommend looking up what are the latest options as these might already be outdated:
Flows: flows.npz and flow consistency masks: flows_con.npz
=> RAFTCameras: refined_cameras.txt
and Depth maps: resized_disps.npz
=> RCVD or newer code for depth/camera estimation from videos, COLMAP is also a reliable starting point to obtain camera poses from the videoPlease refer to the repo's documentation to match the input format our preprocessing code expects, or to adapt it to your own format
Hi, Thank you for the great work! I'm interested in applying your methods to some self-captured videos. Could you please share the process for obtaining the raw data needed for pre-processing? Any guidance would be much appreciated.
Thank you in advance!