Andy Zeng

The code in [demo.cu](https://github.com/andyzeng/3dmatch-toolbox/blob/master/core/demo.cu) uses the point coordinates in your .ply file as well as `voxel_size` (on [line 97](https://github.com/andyzeng/3dmatch-toolbox/blob/master/core/demo.cu#L97)) to automatically estimate how large the TDF voxel volume should be....

Each pose.txt holds a 4x4 [rigid transformation matrix](https://en.wikipedia.org/wiki/Rigid_transformation), which consists of a 3x3 rotation matrix and a 3x1 translation vector: r11, r12, r13, tx r21, r22, r23, ty r31, r32,...

Hello! Have you tried increasing the size of the volumetric voxel grid? With the default parameters ( `voxel_size = 0.006` and `voxel_grid_dim = 500x500x500` ), the voxel grid only expands...

The error occurs because the voxel grid is too large to fit into your memory. Since the code creates two voxel grids (one saving distance values and another saving weights...

> Could this be because the model is in the camera coordinates and it has to be converted to global coordinates? Correct. The fused model created by [demo.cu](https://github.com/andyzeng/tsdf-fusion/blob/master/demo.cu) lies in...

Hello! Yes - one straightforward way would be to construct a voxel grid around your point cloud, where the value of each voxel is the distance to the nearest point...

What is your GCC version? I am using 4.8. Also, I would double check your CUDA 7.5 and cuDNN 5 installation.

Looks like CUDNN_CONVOLUTION_BWD_FILTER_ALGO_WINOGRAD_NONFUSED is new in cuDNN 5.1. Try downloading and installing cuDNN 5.1 and recompiling. I've updated the dependencies information in the [README](https://github.com/andyzeng/3dmatch-toolbox#dependencies).

@caifazhou @JWHennessey Does it work for you if downgrade your GCC version to 4.8? It's possible that CUDA 7.0/7.5 is unhappy with GCC 4.9+.

Hello, it is possible to compute TSDF from a point cloud. However, to make use of negative distance values, you'll need to know "occluded" space in your point cloud. This...