hustvl / 4DGaussians

[CVPR 2024] 4D Gaussian Splatting for Real-Time Dynamic Scene Rendering
https://guanjunwu.github.io/4dgs/
Other
2.05k stars 162 forks source link

Failed generate point clouds from input data using cut_roasted_beef dataset #160

Open raflitp opened 1 month ago

raflitp commented 1 month ago

Hi I'm new to this project, I tried running the given dataset which is the cut_roasted_beef, but I met with an error which doesn't generate any fused points :

Starting fusion with 32 threads Fusing image [1/20] with index 0 in 0.000s (0 points) WARNING: Could not fuse any points. This is likely caused by incorrect settings - filtering must be enabled for the last call to patch match stereo. Number of fused points: 0 Elapsed time: 0.000 [minutes] Writing output: data/dynerf/cut_roasted_beef1/colmap/dense/workspace/fused.ply

I got the fused.ply file but when I tried the third step it doesn't return the points3D_downsample2.ply because the number of fused points are 0. I've tried running using the given example from the readme.md Here I give the screenshot of the full return message. Thanks !!

Screenshot from 2024-07-11 14-49-08

guanjunwu commented 1 month ago

e, what is r_0xx.png? in my colmap process, the generated image will be like this: image

raflitp commented 1 month ago

e, what is r_0xx.png? in my colmap process, the generated image will be like this: image

Not sure what is the r_0xx.png, because it was created from the process. Sort of the images that was split from the video. The result of the first step. How come mine has different output than yours.