Han Hu

Hi, I have figured out a solution for this problem. In `photometric_outlier_detection`, the outlier threshold `gauss_rejection_threshold` is hard coded to 6e-3, which I think is related to the [3-sigma rule](https://en.wikipedia.org/wiki/68%E2%80%9395%E2%80%9399.7_rule)....

same problem on windows. Mac is OK.

For interactive methods, I knew another good work, [OSnap](https://www.cg.tuwien.ac.at/research/publications/2013/arikan-2013-osn/).

Almost forgot, another earlier work, [VideoTrace](http://punchcard.com.au/)

Please find the function fill_hole in [this file](https://github.com/nmoehrle/mvs-texturing/blob/master/libs/tex/generate_texture_patches.cpp) as below. ``` bool fill_hole(std::vector const & hole, UniGraph const & graph, mve::TriangleMesh::ConstPtr mesh, mve::MeshInfo const & mesh_info, std::vector * vertex_projection_infos, std::vector...

Yes, about photogrammetry, but just a job for food.

I also made an implementation and tested for 60 epochs. Reach similar results. The highest is at epoch 58 with IoU of 0.83. ``` Epoch: 58, Test IoU: 0.8301 [10/40]...

Looks good to me~ Just one more thing: one of the beautiful things for RandLA is that, although it's called the selection of random decimation points during the transition-down, in...

I think everyone has his own flavor for coding and I'm fine with this.

I found an issue, the `momentum` parameter in [PyTorch](https://pytorch.org/docs/stable/generated/torch.nn.BatchNorm1d.html) and [Tensorflow](https://www.tensorflow.org/api_docs/python/tf/keras/layers/BatchNormalization) is different. So it should be 0.01 rather than 0.99.