Spaces:
Running
on
Zero
Running
on
Zero
| # https://github.com/yenchenlin/nerf-pytorch/blob/63a5a630c9abd62b0f21c08703d0ac2ea7d4b9dd/run_nerf_helpers.py#L14 | |
| import torch | |
| # torch.autograd.set_detect_anomaly(True) | |
| import torch.nn as nn | |
| import torch.nn.functional as F | |
| import numpy as np | |
| # Misc | |
| img2mse = lambda x, y : torch.mean((x - y) ** 2) | |
| mse2psnr = lambda x : -10. * torch.log(x) / torch.log(torch.Tensor([10.])) | |
| to8b = lambda x : (255*np.clip(x,0,1)).astype(np.uint8) | |
| # Positional encoding (section 5.1) | |
| class Embedder: | |
| def __init__(self, **kwargs): | |
| self.kwargs = kwargs | |
| self.create_embedding_fn() | |
| def create_embedding_fn(self): | |
| embed_fns = [] | |
| d = self.kwargs['input_dims'] | |
| out_dim = 0 | |
| if self.kwargs['include_input']: | |
| embed_fns.append(lambda x : x) | |
| out_dim += d | |
| max_freq = self.kwargs['max_freq_log2'] | |
| N_freqs = self.kwargs['num_freqs'] | |
| if self.kwargs['log_sampling']: | |
| freq_bands = 2.**torch.linspace(0., max_freq, steps=N_freqs) | |
| else: | |
| freq_bands = torch.linspace(2.**0., 2.**max_freq, steps=N_freqs) | |
| for freq in freq_bands: | |
| for p_fn in self.kwargs['periodic_fns']: | |
| embed_fns.append(lambda x, p_fn=p_fn, freq=freq : p_fn(x * freq)) | |
| out_dim += d | |
| self.embed_fns = embed_fns | |
| self.out_dim = out_dim | |
| def embed(self, inputs): | |
| return torch.cat([fn(inputs) for fn in self.embed_fns], -1) | |
| def get_embedder(multires=10, i=0): | |
| if i == -1: | |
| return nn.Identity(), 3 | |
| embed_kwargs = { | |
| 'include_input' : True, | |
| 'input_dims' : 3, | |
| 'max_freq_log2' : multires-1, | |
| 'num_freqs' : multires, | |
| 'log_sampling' : True, | |
| 'periodic_fns' : [torch.sin, torch.cos], | |
| } | |
| embedder_obj = Embedder(**embed_kwargs) | |
| embed = lambda x, eo=embedder_obj : eo.embed(x) | |
| return embed, embedder_obj.out_dim | |