search

prstrive / UniMVSNet

[CVPR 2022] Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation
MIT License
228 stars 12 forks source link

Question about gipuma #19

Open cainsmile opened 2 years ago

cainsmile commented 2 years ago

Hi, you present the point clouds based on gipuma fusion, but I got some problems when I tried to install fusible. Could please me the environment you applied, including the version of CUDA and opencv?

cainsmile commented 2 years ago

I have fixed this problem. However, the number of points I got is only half of the point clouds you provided. Do you where the problem is?

prstrive commented 2 years ago
  1. We adopt the PyTorch 1.2, CUDA 10.0.
  2. Please ensure the fusion parameters (e.g. confidence threshold) are the same with us.
  3. Maybe your fusible install failed due to the mismatch of CUDA or GPU computing power.
  4. If it still doesn't work, Maybe there are something wrong in your dataset format, e.g., camera intrinsics, extrinsics and resolution.
zzy0731 commented 2 years ago

I have same question and i change the compute capability in cmalelist.txt and it works maybe you can try