UniqAI/gaussian-splatting
1
0
Fork 0
mirror of https://github.com/graphdeco-inria/gaussian-splatting synced 2024-11-22 08:18:17 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

chore: prepare for pull-request

Browse Source 
-> removing debug commentaries
-> removing unused proposed code
This commit is contained in:
Stefan Saraev 2024-05-17 17:13:26 +03:00
parent b5a5f72eda
commit 39fb001ef0
5 changed files with 23 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
scene/gaussian_model.py
View File

@ -23,11 +23,11 @@ from utils.general_utils import strip_symmetric, build_scaling_rotation
class GaussianModel:
def setup_functions(self, dtype):
def setup_functions(self):
def build_covariance_from_scaling_rotation(scaling, scaling_modifier, rotation):
L = build_scaling_rotation(scaling_modifier * scaling, rotation, dtype)
L = build_scaling_rotation(scaling_modifier * scaling, rotation)
actual_covariance = L @ L.transpose(1, 2)
symm = strip_symmetric(actual_covariance, dtype)
symm = strip_symmetric(actual_covariance)
return symm
self.scaling_activation = torch.exp
@ -41,7 +41,7 @@ class GaussianModel:
self.rotation_activation = torch.nn.functional.normalize
def __init__(self, sh_degree : int, dtype=torch.float32):
def __init__(self, sh_degree : int):
self.active_sh_degree = 0
self.max_sh_degree = sh_degree
self._xyz = torch.empty(0)
@ -56,8 +56,7 @@ class GaussianModel:
self.optimizer = None
self.percent_dense = 0
self.spatial_lr_scale = 0
self.dtype = dtype
self.setup_functions(dtype)
self.setup_functions()
def capture(self):
return (
@ -137,7 +136,7 @@ class GaussianModel:
rots = torch.zeros((fused_point_cloud.shape[0], 4), device="cuda")
rots[:, 0] = 1
opacities = inverse_sigmoid(0.1 * torch.ones((fused_point_cloud.shape[0], 1), dtype=self.dtype, device="cuda"))
opacities = inverse_sigmoid(0.1 * torch.ones((fused_point_cloud.shape[0], 1), dtype=torch.float, device="cuda"))
self._xyz = nn.Parameter(fused_point_cloud.requires_grad_(True))
self._features_dc = nn.Parameter(features[:,:,0:1].transpose(1, 2).contiguous().requires_grad_(True))
@ -145,7 +144,7 @@ class GaussianModel:
self._scaling = nn.Parameter(scales.requires_grad_(True))
self._rotation = nn.Parameter(rots.requires_grad_(True))
self._opacity = nn.Parameter(opacities.requires_grad_(True))
self.max_radii2D = torch.zeros((self.get_xyz.shape[0]), device="cuda", dtype=self.dtype)
self.max_radii2D = torch.zeros((self.get_xyz.shape[0]), device="cuda")
def training_setup(self, training_args):
self.percent_dense = training_args.percent_dense
@ -247,12 +246,12 @@ class GaussianModel:
for idx, attr_name in enumerate(rot_names):
rots[:, idx] = np.asarray(plydata.elements[0][attr_name])
self._xyz = nn.Parameter(torch.tensor(xyz, dtype=self.dtype, device="cuda").requires_grad_(True))
self._features_dc = nn.Parameter(torch.tensor(features_dc, dtype=self.dtype, device="cuda").transpose(1, 2).contiguous().requires_grad_(True))
self._features_rest = nn.Parameter(torch.tensor(features_extra, dtype=self.dtype, device="cuda").transpose(1, 2).contiguous().requires_grad_(True))
self._opacity = nn.Parameter(torch.tensor(opacities, dtype=self.dtype, device="cuda").requires_grad_(True))
self._scaling = nn.Parameter(torch.tensor(scales, dtype=self.dtype, device="cuda").requires_grad_(True))
self._rotation = nn.Parameter(torch.tensor(rots, dtype=self.dtype, device="cuda").requires_grad_(True))
self._xyz = nn.Parameter(torch.tensor(xyz, dtype=torch.float, device="cuda").requires_grad_(True))
self._features_dc = nn.Parameter(torch.tensor(features_dc, dtype=torch.float, device="cuda").transpose(1, 2).contiguous().requires_grad_(True))
self._features_rest = nn.Parameter(torch.tensor(features_extra, dtype=torch.float, device="cuda").transpose(1, 2).contiguous().requires_grad_(True))
self._opacity = nn.Parameter(torch.tensor(opacities, dtype=torch.float, device="cuda").requires_grad_(True))
self._scaling = nn.Parameter(torch.tensor(scales, dtype=torch.float, device="cuda").requires_grad_(True))
self._rotation = nn.Parameter(torch.tensor(rots, dtype=torch.float, device="cuda").requires_grad_(True))
self.active_sh_degree = self.max_sh_degree

3
train.py
View File

@ -16,7 +16,7 @@ from utils.loss_utils import l1_loss, ssim
from gaussian_renderer import render, network_gui
import sys
from scene import Scene, GaussianModel
from utils.general_utils import get_data_dtype, safe_state
from utils.general_utils import safe_state
import uuid
from tqdm import tqdm
from utils.image_utils import psnr
@ -216,7 +216,6 @@ if __name__ == "__main__":
# Start GUI server, configure and run training
network_gui.init(args.ip, args.port)
torch.autograd.set_detect_anomaly(args.detect_anomaly)
training(lp.extract(args), op.extract(args), pp.extract(args), args.test_iterations, args.save_iterations, args.checkpoint_iterations, args.start_checkpoint, args.debug_from)
# All done

2
utils/camera_utils.py
View File

@ -40,8 +40,6 @@ def loadCam(args, id, cam_info, resolution_scale):
resized_image_rgb = PILtoTorch(cam_info.image, resolution)
# resized_image_rgb = resized_image_rgb.to(get_data_dtype(args.data_dtype))
gt_image = resized_image_rgb[:3, ...]
loaded_mask = None

14
utils/general_utils.py
View File

@ -20,7 +20,7 @@ def inverse_sigmoid(x):
def PILtoTorch(pil_image, resolution):
resized_image_PIL = pil_image.resize(resolution)
resized_image = torch.from_numpy(np.array(resized_image_PIL))# / 255.0
resized_image = torch.from_numpy(np.array(resized_image_PIL))
if len(resized_image.shape) == 3:
return resized_image.permute(2, 0, 1)
else:
@ -61,8 +61,8 @@ def get_expon_lr_func(
return helper
def strip_lowerdiag(L, dtype=torch.float32):
uncertainty = torch.zeros((L.shape[0], 6), dtype=dtype, device="cuda")
def strip_lowerdiag(L):
uncertainty = torch.zeros((L.shape[0], 6), dtype=torch.float, device="cuda")
uncertainty[:, 0] = L[:, 0, 0]
uncertainty[:, 1] = L[:, 0, 1]
@ -72,8 +72,8 @@ def strip_lowerdiag(L, dtype=torch.float32):
uncertainty[:, 5] = L[:, 2, 2]
return uncertainty
def strip_symmetric(sym, dtype=torch.float32):
return strip_lowerdiag(sym, dtype=dtype)
def strip_symmetric(sym):
return strip_lowerdiag(sym)
def build_rotation(r):
norm = torch.sqrt(r[:,0]*r[:,0] + r[:,1]*r[:,1] + r[:,2]*r[:,2] + r[:,3]*r[:,3])
@ -98,8 +98,8 @@ def build_rotation(r):
R[:, 2, 2] = 1 - 2 * (x*x + y*y)
return R
def build_scaling_rotation(s, r, dtype=torch.float32):
L = torch.zeros((s.shape[0], 3, 3), dtype=dtype, device="cuda")
def build_scaling_rotation(s, r):
L = torch.zeros((s.shape[0], 3, 3), dtype=torch.float, device="cuda")
R = build_rotation(r)
L[:,0,0] = s[:,0]