Example 3: Video Generation from Multiview Images

[Xchanger0321]

Hello,
Thanks for the amazing repo.

I have a doubt in generating npz (numpy) file for GEN3C Multiview.

"These can be obtained using off-the-shelf methods like VGGT."
I Followed the above repo and generated npy files from VGTT , however the file names are different from the dataset from hugging face.

VGGT NPZ : depth.npy, depth_conf.npy , extrinsic.npy , images.npy , intrinsic.npy , pose_enc.npy , pose_enc_list.npy ,world_points.npy , world_points_conf.npy , world_points_from_depth.npy

GEN3C Dataset: depth_key_frames.npy , images_key_frames.npy , K_key_frames.npy , Ks_all.npy , mask_key_frames.npy , w2cs_all.npy , w2cs_key_frames.npy

can anyone guide me on generating NPZ file data set from VGTT and feed it to Multiview GEN3C?

[cerebrai]

Same issue .. would be great if there is a script or some help on how to convert the output of vggt to become compatible with GEN3C multiview code.

[Xchanger0321]

@cerebrai , follow this repo to use custom dataset for multiview

https://github.com/dsvilarkovic/GEN3C/blob/main/README_MULTIVIEW.md

while I'm creating a script to convert VGGT NPZ to GEN3C format , you can meanwhile try the above repo

[Ibrahimshoer93]

Any updates on this?