Open graveyard99 opened 6 hours ago
Hi. The script of video reconstruction, which is an extension of MoGe, is not released yet. We have currently released only the Monocular Geometry estimation inference code, which is the original purpose of MoGe.
These two issues might be related to your need, where you can find how we implemented the video reconstruction code.
Kind of useful but not exactly what creatives need if they want to accurately add object to a scene with matching perspective/parallax
Are you familiar with software like MetaShape, RealityCapture or PFTrack?
These all employ photogrammetric SFM algorithms to find camera positions and enable the export of all aligned camera positions in a wide range of file formats for VFX artists to derive camera tracks from
Kind Regards
Hayden Martin https://www.haydenmart.in/
Instagram https://www.instagram.com/_hayden_martin_/ Vimeo https://vimeo.com/haydenmartin LinkedIn https://uk.linkedin.com/in/haydenmartincreative +44 07817281488
On Tue, Nov 5, 2024 at 3:31 PM Ruicheng Wang @.***> wrote:
Hi. The script of video reconstruction, which is an extension of MoGe, is not released yet. We have currently released only the Monocular Geometry estimation inference code, which is the original purpose of MoGe.
These two issues might be related to your need, where you can find how we implemented the video reconstruction code.
— Reply to this email directly, view it on GitHub https://github.com/microsoft/MoGe/issues/12#issuecomment-2457483694, or unsubscribe https://github.com/notifications/unsubscribe-auth/AL24KGHUV7OE6IUD4VCSZFLZ7DQEPAVCNFSM6AAAAABRGXSVUGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINJXGQ4DGNRZGQ . You are receiving this because you authored the thread.Message ID: @.***>
Thank you for your suggestions. We will consider and explore more about the community needs in our future research.
If you are essentially registering each frame of the point cloud to create the static environment, are you also outputting the camera positions?
if there's no animated camera data it's not that useful.
FBX and Alembic formats would be great for this