mirror of
https://github.com/facebookresearch/pytorch3d.git
synced 2025-08-01 03:12:49 +08:00
29b8ebd802
Summary: Introduce methods to approximate the radii of conical frustums along rays as described in [MipNerf](https://arxiv.org/abs/2103.13415): - Two new attributes are added to ImplicitronRayBundle: bins and radii. Bins is of size n_pts_per_ray + 1. It allows us to manipulate easily and n_pts_per_ray intervals. For example we need the intervals coordinates in the radii computation for \(t_{\mu}, t_{\delta}\). Radii are used to store the radii of the conical frustums. - Add 3 new methods to compute the radii: - approximate_conical_frustum_as_gaussians: It computes the mean along the ray direction, the variance of the conical frustum with respect to t and variance of the conical frustum with respect to its radius. This implementation follows the stable computation defined in the paper. - compute_3d_diagonal_covariance_gaussian: Will leverage the two previously computed variances to find the diagonal covariance of the Gaussian. - conical_frustum_to_gaussian: Mix everything together to compute the means and the diagonal covariances along the ray of the Gaussians. - In AbstractMaskRaySampler, introduces the attribute `cast_ray_bundle_as_cone`. If False it won't change the previous behaviour of the RaySampler. However if True, the samplers will sample `n_pts_per_ray +1` instead of `n_pts_per_ray`. This points are then used to set the bins attribute of ImplicitronRayBundle. The support of HeterogeneousRayBundle has not been added since the current code does not allow it. A safeguard has been added to avoid a silent bug in the future. Reviewed By: shapovalov Differential Revision: D45269190 fbshipit-source-id: bf22fad12d71d55392f054e3f680013aa0d59b78
127 lines
2.9 KiB
YAML
127 lines
2.9 KiB
YAML
log_vars:
|
|
- loss_rgb_psnr_fg
|
|
- loss_rgb_psnr
|
|
- loss_rgb_mse
|
|
- loss_rgb_huber
|
|
- loss_depth_abs
|
|
- loss_depth_abs_fg
|
|
- loss_mask_neg_iou
|
|
- loss_mask_bce
|
|
- loss_mask_beta_prior
|
|
- loss_eikonal
|
|
- loss_density_tv
|
|
- loss_depth_neg_penalty
|
|
- loss_autodecoder_norm
|
|
- loss_prev_stage_rgb_mse
|
|
- loss_prev_stage_rgb_psnr_fg
|
|
- loss_prev_stage_rgb_psnr
|
|
- loss_prev_stage_mask_bce
|
|
- objective
|
|
- epoch
|
|
- sec/it
|
|
mask_images: true
|
|
mask_depths: true
|
|
render_image_width: 400
|
|
render_image_height: 400
|
|
mask_threshold: 0.5
|
|
output_rasterized_mc: false
|
|
bg_color:
|
|
- 0.0
|
|
- 0.0
|
|
- 0.0
|
|
num_passes: 1
|
|
chunk_size_grid: 4096
|
|
render_features_dimensions: 3
|
|
tqdm_trigger_threshold: 16
|
|
n_train_target_views: 1
|
|
sampling_mode_training: mask_sample
|
|
sampling_mode_evaluation: full_grid
|
|
global_encoder_class_type: SequenceAutodecoder
|
|
raysampler_class_type: AdaptiveRaySampler
|
|
renderer_class_type: LSTMRenderer
|
|
image_feature_extractor_class_type: ResNetFeatureExtractor
|
|
view_pooler_enabled: true
|
|
implicit_function_class_type: IdrFeatureField
|
|
view_metrics_class_type: ViewMetrics
|
|
regularization_metrics_class_type: RegularizationMetrics
|
|
loss_weights:
|
|
loss_rgb_mse: 1.0
|
|
loss_prev_stage_rgb_mse: 1.0
|
|
loss_mask_bce: 0.0
|
|
loss_prev_stage_mask_bce: 0.0
|
|
global_encoder_SequenceAutodecoder_args:
|
|
autodecoder_args:
|
|
encoding_dim: 0
|
|
n_instances: 1
|
|
init_scale: 1.0
|
|
ignore_input: false
|
|
raysampler_AdaptiveRaySampler_args:
|
|
n_pts_per_ray_training: 64
|
|
n_pts_per_ray_evaluation: 64
|
|
n_rays_per_image_sampled_from_mask: 1024
|
|
n_rays_total_training: null
|
|
stratified_point_sampling_training: true
|
|
stratified_point_sampling_evaluation: false
|
|
cast_ray_bundle_as_cone: false
|
|
scene_extent: 8.0
|
|
scene_center:
|
|
- 0.0
|
|
- 0.0
|
|
- 0.0
|
|
renderer_LSTMRenderer_args:
|
|
num_raymarch_steps: 10
|
|
init_depth: 17.0
|
|
init_depth_noise_std: 0.0005
|
|
hidden_size: 16
|
|
n_feature_channels: 256
|
|
bg_color: null
|
|
verbose: false
|
|
image_feature_extractor_ResNetFeatureExtractor_args:
|
|
name: resnet34
|
|
pretrained: true
|
|
stages:
|
|
- 1
|
|
- 2
|
|
- 3
|
|
- 4
|
|
normalize_image: true
|
|
image_rescale: 0.16
|
|
first_max_pool: true
|
|
proj_dim: 32
|
|
l2_norm: true
|
|
add_masks: true
|
|
add_images: true
|
|
global_average_pool: false
|
|
feature_rescale: 1.0
|
|
view_pooler_args:
|
|
feature_aggregator_class_type: AngleWeightedIdentityFeatureAggregator
|
|
view_sampler_args:
|
|
masked_sampling: false
|
|
sampling_mode: bilinear
|
|
feature_aggregator_AngleWeightedIdentityFeatureAggregator_args:
|
|
exclude_target_view: true
|
|
exclude_target_view_mask_features: true
|
|
concatenate_output: true
|
|
weight_by_ray_angle_gamma: 1.0
|
|
min_ray_angle_weight: 0.1
|
|
implicit_function_IdrFeatureField_args:
|
|
d_in: 3
|
|
d_out: 1
|
|
dims:
|
|
- 512
|
|
- 512
|
|
- 512
|
|
- 512
|
|
- 512
|
|
- 512
|
|
- 512
|
|
- 512
|
|
geometric_init: true
|
|
bias: 1.0
|
|
skip_in: []
|
|
weight_norm: true
|
|
n_harmonic_functions_xyz: 1729
|
|
pooled_feature_dim: 0
|
|
view_metrics_ViewMetrics_args: {}
|
|
regularization_metrics_RegularizationMetrics_args: {}
|