sergeyprokudin

# following
# https://github.com/facebookresearch/pytorch3d/issues/35
import torch
import torch.nn as nn

class PointsRendererDepth(nn.Module):
    """
    A class for rendering a batch of points. The class should
    be initialized with a rasterizer and compositor class which each have a forward
    function.

sergeyprokudin

"""
MIT License

Copyright (c) 2024 Sergey Prokudin [email protected]

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is

sergeyprokudin

# Pymeshlab normal estimation (and optional Poisson)

# https://github.com/cnr-isti-vclab/PyMeshLab/blob/main/docs/filter_list.rst
# https://pymeshlab.readthedocs.io/en/0.2/tutorials/get_mesh_values.html

import pymeshlab

def get_normals_pymeshlab(x,  k=10, smoothiter=0):

  tmp_mesh_path = 'tmp.obj'

sergeyprokudin

import torch
import torch.nn as nn

def laplacian_smoothing_loss2d(Y_pred):
    C, H, W = Y_pred.shape[1], Y_pred.shape[2], Y_pred.shape[3]
    kernel = torch.tensor([[0, 1, 0], [1, -4, 1], [0, 1, 0]], device=device, dtype=torch.float32)
    kernel = kernel.view(1, 1, 3, 3).repeat(C, 1, 1, 1)
    Y_laplacian = nn.functional.conv2d(Y_pred, kernel, groups=C, padding=1)
    laplacian_loss = (Y_laplacian ** 2).mean()
    return laplacian_loss

sergeyprokudin

# simple positional embedding (NeRF style)
# as in https://arxiv.org/pdf/2003.08934 (Equation 4)

def positional_embedding(x, k):

    batch_size, n_dims = x.shape
    x_rep = x.unsqueeze(-1).expand(-1, -1, k)

    k_pows = torch.pow(2, torch.arange(0, k, device=x.device)).view(1, 1, k)
    x_k = k_pows * torch.pi * x_rep

sergeyprokudin

# @title Define SIREN deformation model

# https://github.com/vsitzmann/siren

# MIT License

# Copyright (c) 2020 Vincent Sitzmann

# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal