Files
nerfstudio/nerfactory/models/instant_ngp.py
T

258 lines
9.3 KiB
Python

# Copyright 2022 The Plenoptix Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Implementation of Instant NGP.
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import Dict, List, Optional, Tuple, Type
import nerfacc
import torch
from nerfacc import ContractionType
from torch.nn import Parameter
from torchmetrics import PeakSignalNoiseRatio
from torchmetrics.functional import structural_similarity_index_measure
from torchmetrics.image.lpip import LearnedPerceptualImagePatchSimilarity
from nerfactory.cameras.rays import RayBundle
from nerfactory.fields.instant_ngp_field import TCNNInstantNGPField
from nerfactory.fields.modules.field_heads import FieldHeadNames
from nerfactory.models.base import Model, ModelConfig
from nerfactory.models.modules.ray_sampler import VolumetricSampler
from nerfactory.optimizers.loss import MSELoss
from nerfactory.renderers.renderers import (
AccumulationRenderer,
DepthRenderer,
RGBRenderer,
)
from nerfactory.utils import colors, visualization
from nerfactory.utils.callbacks import (
TrainingCallback,
TrainingCallbackAttributes,
TrainingCallbackLocation,
)
@dataclass
class InstantNGPModelConfig(ModelConfig):
"""Instant NGP Model Config"""
_target: Type = field(
default_factory=lambda: NGPModel
) # We can't write `NGPModel` directly, because `NGPModel` doesn't exist yet
"""target class to instantiate"""
enable_collider: bool = False
"""Whether to create a scene collider to filter rays."""
collider_params: Optional[Dict[str, float]] = None
"""Instant NGP doesn't use a collider."""
max_num_samples_per_ray: int = 24
"""Number of samples in field evaluation."""
grid_resolution: int = 128
"""Resolution of the grid used for the field."""
contraction_type: ContractionType = ContractionType.UN_BOUNDED_SPHERE
"""Resolution of the grid used for the field."""
cone_angle: float = 0.004
"""Should be set to 0.0 for blender scenes but 1./256 for real scenes."""
render_step_size: float = 0.01
"""Minimum step size for rendering."""
near_plane: float = 0.05
"""How far along ray to start sampling."""
far_plane: float = 1e3
"""How far along ray to stop sampling."""
use_appearance_embedding: bool = False
"""Whether to use an appearance embedding."""
randomize_background: bool = True
"""Whether to randomize the background color."""
class NGPModel(Model):
"""Instant NGP model
Args:
config: instant NGP configuration to instantiate model
"""
config: InstantNGPModelConfig
field: TCNNInstantNGPField
def __init__(self, config: InstantNGPModelConfig, **kwargs) -> None:
super().__init__(config=config, **kwargs)
def populate_modules(self):
"""Set the fields and modules."""
super().populate_modules()
self.field = TCNNInstantNGPField(
aabb=self.scene_bounds.aabb,
contraction_type=self.config.contraction_type,
use_appearance_embedding=self.config.use_appearance_embedding,
num_images=self.num_train_data,
)
self.scene_aabb = Parameter(self.scene_bounds.aabb.flatten(), requires_grad=False)
# Occupancy Grid
self.occupancy_grid = nerfacc.OccupancyGrid(
roi_aabb=self.scene_aabb,
resolution=self.config.grid_resolution,
contraction_type=self.config.contraction_type,
)
# Sampler
vol_sampler_aabb = self.scene_bounds.aabb if self.config.contraction_type == ContractionType.AABB else None
self.sampler = VolumetricSampler(
scene_aabb=vol_sampler_aabb,
occupancy_grid=self.occupancy_grid,
density_fn=self.field.density_fn,
)
# renderers
background_color = None if self.config.randomize_background else colors.WHITE
self.renderer_rgb = RGBRenderer(background_color=background_color)
self.renderer_accumulation = AccumulationRenderer()
self.renderer_depth = DepthRenderer()
# losses
self.rgb_loss = MSELoss()
# metrics
self.psnr = PeakSignalNoiseRatio(data_range=1.0)
self.ssim = structural_similarity_index_measure
self.lpips = LearnedPerceptualImagePatchSimilarity()
def get_training_callbacks(
self, training_callback_attributes: TrainingCallbackAttributes
) -> List[TrainingCallback]:
def update_occupancy_grid(step: int):
self.occupancy_grid.every_n_step(
step=step,
occ_eval_fn=lambda x: self.field.get_opacity(x, self.config.render_step_size),
)
return [
TrainingCallback(
where_to_run=[TrainingCallbackLocation.BEFORE_TRAIN_ITERATION],
update_every_num_iters=1,
func=update_occupancy_grid,
),
]
def get_param_groups(self) -> Dict[str, List[Parameter]]:
param_groups = {}
if self.field is None:
raise ValueError("populate_fields() must be called before get_param_groups")
param_groups["fields"] = list(self.field.parameters())
return param_groups
def get_outputs(self, ray_bundle: RayBundle):
assert self.field is not None
num_rays = len(ray_bundle)
with torch.no_grad():
ray_samples, packed_info, ray_indices = self.sampler(
ray_bundle=ray_bundle,
near_plane=self.config.near_plane,
far_plane=self.config.far_plane,
render_step_size=self.config.render_step_size,
cone_angle=self.config.cone_angle,
)
field_outputs = self.field(ray_samples)
# accumulation
weights = nerfacc.render_weight_from_density(
packed_info=packed_info,
sigmas=field_outputs[FieldHeadNames.DENSITY],
t_starts=ray_samples.frustums.starts,
t_ends=ray_samples.frustums.ends,
)
rgb = self.renderer_rgb(
rgb=field_outputs[FieldHeadNames.RGB],
weights=weights,
ray_indices=ray_indices,
num_rays=num_rays,
)
depth = self.renderer_depth(
weights=weights, ray_samples=ray_samples, ray_indices=ray_indices, num_rays=num_rays
)
accumulation = self.renderer_accumulation(weights=weights, ray_indices=ray_indices, num_rays=num_rays)
alive_ray_mask = accumulation.squeeze(-1) > 0
outputs = {
"rgb": rgb,
"accumulation": accumulation,
"depth": depth,
"alive_ray_mask": alive_ray_mask, # the rays we kept from sampler
"num_samples_per_ray": packed_info[:, 1],
}
return outputs
def get_metrics_dict(self, outputs, batch):
image = batch["image"].to(self.device)
metrics_dict = {}
metrics_dict["psnr"] = self.psnr(outputs["rgb"], image)
metrics_dict["num_samples_per_batch"] = outputs["num_samples_per_ray"].sum()
return metrics_dict
def get_loss_dict(self, outputs, batch, metrics_dict=None):
image = batch["image"].to(self.device)
mask = outputs["alive_ray_mask"]
rgb_loss = self.rgb_loss(image[mask], outputs["rgb"][mask])
loss_dict = {"rgb_loss": rgb_loss}
return loss_dict
def get_image_metrics_and_images(
self, outputs: Dict[str, torch.Tensor], batch: Dict[str, torch.Tensor]
) -> Tuple[Dict[str, float], Dict[str, torch.Tensor]]:
image = batch["image"].to(self.device)
rgb = outputs["rgb"]
acc = visualization.apply_colormap(outputs["accumulation"])
depth = visualization.apply_depth_colormap(
outputs["depth"],
accumulation=outputs["accumulation"],
)
alive_ray_mask = visualization.apply_colormap(outputs["alive_ray_mask"])
combined_rgb = torch.cat([image, rgb], dim=1)
combined_acc = torch.cat([acc], dim=1)
combined_depth = torch.cat([depth], dim=1)
combined_alive_ray_mask = torch.cat([alive_ray_mask], dim=1)
# Switch images from [H, W, C] to [1, C, H, W] for metrics computations
image = torch.moveaxis(image, -1, 0)[None, ...]
rgb = torch.moveaxis(rgb, -1, 0)[None, ...]
psnr = self.psnr(image, rgb)
ssim = self.ssim(image, rgb)
lpips = self.lpips(image, rgb)
# all of these metrics will be logged as scalars
metrics_dict = {"psnr": float(psnr.item()), "ssim": float(ssim), "lpips": float(lpips)} # type: ignore
# TODO(ethan): return an image dictionary
images_dict = {
"img": combined_rgb,
"accumulation": combined_acc,
"depth": combined_depth,
"alive_ray_mask": combined_alive_ray_mask,
}
return metrics_dict, images_dict