search

SJoJoK / 3DGStream

[CVPR 2024 Highlight] Official repository for the paper "3DGStream: On-the-fly Training of 3D Gaussians for Efficient Streaming of Photo-Realistic Free-Viewpoint Videos".
https://sjojok.github.io/3dgstream
MIT License
343 stars 24 forks source link

there is no training process. #21

Open krystalonly opened 4 months ago

krystalonly commented 4 months ago

1 My environment is from gaussian_splatting. And I use flame_steak in text file directly.

liysu commented 4 months ago

I have the same problem. Have you solved it？

MarcWangzhiru commented 4 months ago

You need to make your own dataset, only the first frame and trained point cloud are provided in the test folder, you need to extract the subsequent frames yourself and organize them as demo. In fact if you print (frames) you'll get an empty list in train_frames.py. 屏幕截图 2024-07-11 010733

SJoJoK commented 4 months ago

Yeah, just as @MarcWangzhiru said, you need to follow the README and organize the multi-view dataset by yourself.