Spaces:
Sleeping
Sleeping
File size: 31,907 Bytes
801501a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 |
from __future__ import annotations
import constants
import vtk
import vtk.util.numpy_support as numpy_support
from custom_types import *
from utils import files_utils, rotation_utils
from models import gm_utils
from ui import ui_utils, inference_processing, gaussian_status
import options
def filter_by_inclusion(gaussian: gaussian_status.GaussianStatus) -> bool:
return gaussian.included
def filter_by_selection(gaussian: gaussian_status.GaussianStatus) -> bool:
return gaussian.is_selected
class GmmMeshStage:
def turn_off_selected(self):
if self.selected is not None:
# self.arrows.turn_off()
self.toggle_selection(self.selected)
self.selected = None
def turn_gmm_off(self):
self.turn_off_selected()
for gaussian in self.gmm:
gaussian.turn_off()
def turn_gmm_on(self):
for gaussian in self.gmm:
gaussian.turn_on()
def event_manger(self, object_id: str):
if object_id in self.addresses_dict:
return self.toggle_selection(object_id)
elif self.arrows.check_event(object_id):
transform = self.arrows.get_transform(object_id)
self.update_gmm(*transform)
return True
return False
def toggle_selection(self, object_id: str):
self.gmm[self.addresses_dict[object_id]].toggle_selection()
if self.selected is None:
self.selected = object_id
elif self.selected == object_id and self.gmm[self.addresses_dict[object_id]].is_not_selected:
self.selected = None
else:
self.gmm[self.addresses_dict[self.selected]].toggle_selection()
self.selected = object_id
# if self.selected is not None:
# self.arrows.update_arrows_transform(self.gmm[self.addresses_dict[self.selected]])
# else:
# self.arrows.turn_off()
return True
def toggle_inclusion_by_id(self, g_id: int, select: Optional[bool] = None) -> Tuple[bool, List[gaussian_status.GaussianStatus]]:
toggled = []
self.gmm[g_id].toggle_inclusion(select)
toggled.append(self.gmm[g_id])
if self.symmetric_mode:
if self.gmm[g_id].twin is not None and self.gmm[g_id].twin.included != self.gmm[g_id].included:
self.gmm[g_id].twin.toggle_inclusion(select)
toggled.append(self.gmm[g_id].twin)
return True, toggled
def toggle_inclusion(self, object_id: str) -> Tuple[bool, List[gaussian_status.GaussianStatus]]:
if object_id in self.addresses_dict:
return self.toggle_inclusion_by_id(self.addresses_dict[object_id])
return False, []
def toggle_all(self):
for gaussian in self.gmm:
gaussian.toggle_inclusion()
def __len__(self):
return len(self.gmm)
def set_opacity(self, opacity: float):
self.view_style.opacity = opacity
for gaussian in self.gmm:
gaussian.set_color()
def update_gmm(self, button: ui_utils.Buttons, key: str) -> bool:
if self.selected is not None:
g_id = self.addresses_dict[self.selected]
self.gmm[g_id].apply_affine(button, key)
if self.symmetric_mode:
if self.gmm[g_id].twin is not None:
self.gmm[g_id].twin.make_symmetric(False)
# self.arrows.update_arrows_transform(self.gmm[self.addresses_dict[self.selected]])
return True
return False
def get_gmm(self) -> Tuple[TS, T]:
raw_gmm = [g.get_raw_data() for g in self.gmm if g.included]
phi = torch.tensor([g[0] for g in raw_gmm], dtype=torch.float32).view(1, 1, -1)
# phi = torch.from_numpy(self.raw_gmm[0]).view(1, 1, -1).float()
mu = torch.stack([torch.from_numpy(g[1]).float() for g in raw_gmm], dim=0).view(1, 1, -1, 3)
p = torch.stack([torch.from_numpy(g[3]).float() for g in raw_gmm], dim=0).view(1, 1, -1, 3, 3)
eigen = torch.stack([torch.from_numpy(g[2]).float() for g in raw_gmm], dim=0).view(1, 1, -1, 3)
gmm = mu, p, phi, eigen
included = torch.tensor([g.gaussian_id for g in self.gmm if g.included], dtype=torch.int64)
return gmm, included
def reset(self):
for g in self.gmm:
g.reset()
# self.turn_off_selected()
def remove_all(self):
self.remove_gaussians(list(self.addresses_dict.keys()))
self.addresses_dict = {}
self.gmm = []
# def switch_arrows(self, arrow_type: ui_utils.Buttons):
# if self.arrows.switch_arrows(arrow_type) and self.selected is not None:
# self.arrows.update_arrows_transform(self.gmm[self.addresses_dict[self.selected]])
def toggle_symmetric(self, force_include: bool):
self.symmetric_mode = not self.symmetric_mode and False
# visited = set()
if self.symmetric_mode:
for i in range(len(self)):
self.gmm[i].make_symmetric(force_include)
def remove_gaussians(self, addresses: List[str]):
for address in addresses:
gaussian_id: int = self.addresses_dict[address]
gaussian = self.gmm[gaussian_id]
# if gaussian.is_selected:
# self.toggle_selection(address)
self.gmm[gaussian_id] = None
gaussian.delete(self.render)
del self.addresses_dict[address]
self.gmm = [gaussian for gaussian in self.gmm if gaussian is not None]
self.addresses_dict = {self.gmm[i].get_address(): i for i in range(len(self.gmm))}
def add_gaussians(self, gaussians: List[gaussian_status.GaussianStatus]) -> List[str]:
new_addresses = []
for i, gaussian in enumerate(gaussians):
gaussian_copy = gaussian.copy(self.render, self.view_style, is_selected=False)
self.gmm.append(gaussian_copy)
new_addresses.append(gaussian_copy.get_address())
self.addresses_dict = {self.gmm[i].get_address(): i for i in range(len(self.gmm))}
return new_addresses
def make_twins(self, address_a: str, address_b: str):
if address_a in self.addresses_dict and address_b in self.addresses_dict:
gaussian_a, gaussian_b = self.gmm[self.addresses_dict[address_a]], self.gmm[self.addresses_dict[address_b]]
gaussian_a.twin = gaussian_b
gaussian_b.twin = gaussian_a
def split_mesh_by_gmm(self, mesh) -> Dict[int, T]:
faces_split = {}
mu, p, phi, _ = self.get_gmm()[0]
eigen = torch.stack([torch.from_numpy(g.get_view_eigen()).float() for g in self.gmm if g.included], dim=0).view(1, 1, -1, 3)
gmm = mu, p, phi, eigen
faces_split_ = gm_utils.split_mesh_by_gmm(mesh, gmm)
counter = 0
for i in range(len(self.gmm)):
if self.gmm[i].disabled:
faces_split[i] = None
else:
faces_split[i] = faces_split_[counter]
counter += 1
return faces_split
@staticmethod
def get_part_face(mesh: V_Mesh, faces_inds: T) -> Tuple[T_Mesh, T]:
mesh = mesh[0], torch.from_numpy(mesh[1]).long()
mask = faces_inds.ne(0)
faces = mesh[1][mask]
vs_inds = faces.flatten().unique()
vs = mesh[0][vs_inds]
mapper = torch.zeros(mesh[0].shape[0], dtype=torch.int64)
mapper[vs_inds] = torch.arange(vs.shape[0])
return (vs, mapper[faces]), faces_inds[mask]
def save(self, root: str, filter_faces: Callable[[gaussian_status.GaussianStatus], bool] = filter_by_inclusion):
if self.faces is not None:
if self.gmm_id == -1:
name = "mix"
else:
name = str(self.gmm_id)
path = f"{root}/{files_utils.get_time_name(name)}"
faces = list(filter(lambda x: x[1] is not None, self.faces.items()))
mesh = self.vs, np.concatenate(list(map(lambda x: x[1], faces)))
faces_inds = map(lambda x:
torch.ones(x[1].shape[0], dtype=torch.int64)
if filter_faces(self.gmm[x[0]]) else torch.zeros(x[1].shape[0], dtype=torch.int64), faces)
faces_inds = torch.cat(list(faces_inds))
# if name != 'mix':
# mesh, faces_inds = self.get_part_face(mesh, faces_inds)
files_utils.export_mesh(mesh, path)
files_utils.export_list(faces_inds.tolist(), f"{path}_faces")
def aggregate_symmetric(self) -> Dict[str, int]:
if not self.symmetric_mode:
return self.votes
out = {}
for item in self.votes:
actor_id = self.addresses_dict[item]
twin = self.gmm[actor_id].twin
out[item] = self.votes[item]
if twin is not None and twin.get_address() not in self.votes:
out[twin.get_address()] = self.votes[item]
return out
def aggregate_votes(self) -> List[int]:
# to_do = self.add_selection if select else self.clear_selection
actors_id = []
# votes = self.aggregate_symmetric()
for item in self.votes:
actor_id = self.addresses_dict[item]
actors_id.append(actor_id)
self.votes = {}
return actors_id
def vote(self, *actors: Optional[vtk.vtkActor]):
for actor in actors:
if actor is not None:
address = actor.GetAddressAsString('')
if address in self.addresses_dict:
if address not in self.votes:
self.votes[address] = 0
self.votes[address] += 1
@staticmethod
def faces_to_vtk_faces(faces: Union[T, ARRAY]):
if type(faces) is T:
faces = faces.detach().cpu().numpy()
cells_npy = np.column_stack(
[np.full(faces.shape[0], 3, dtype=np.int64), faces.astype(np.int64)]).ravel()
faces_vtk = vtk.vtkCellArray()
faces_vtk.SetCells(faces.shape[0], numpy_support.numpy_to_vtkIdTypeArray(cells_npy))
return faces_vtk
def get_mesh_part(self, vs: vtk.vtkPoints, faces: Optional[Union[T, ARRAY]]) -> Optional[vtk.vtkPolyData]:
if faces is not None:
# actor_mesh = vtk.vtkActor()
mesh = vtk.vtkPolyData()
# mapper = vtk.vtkPolyDataMapper()
mesh.SetPoints(vs)
mesh.SetPolys(self.faces_to_vtk_faces(faces))
# mapper.SetInputData(mesh)
# actor_mesh.SetMapper(mapper)
# actor_mesh.GetProperty().SetOpacity(0.3)
# actor_mesh.PickableOff()
# if self.to_init:
# self.render.AddActor(actor_mesh)
return mesh
return None
def add_gmm(self) -> List[gaussian_status.GaussianStatus]:
gmms = []
if len(self.raw_gmm) > 0:
phi = self.raw_gmm[0]
phi = np.exp(phi)
phi = phi / phi.sum()
for i, gaussian in enumerate(zip(*self.raw_gmm)):
gaussian = gaussian_status.GaussianStatus(gaussian, (self.gmm_id, i), False, self.view_style,
self.render, phi[i])
gmms.append(gaussian)
return gmms
def add_mesh(self, base_mesh: T_Mesh, split_mesh: bool = True, for_slider: bool = True):
if base_mesh is not None:
vs_vtk = vtk.vtkPoints()
self.vs = base_mesh[0]
if for_slider:
vs_ui = self.init_mesh_pos(base_mesh[0])
else:
vs_ui = self.vs
vs_vtk.SetData(numpy_support.numpy_to_vtk(vs_ui.numpy()))
if split_mesh:
self.faces = self.split_mesh_by_gmm(base_mesh)
for i in range(len(self.gmm)):
part_mesh = self.get_mesh_part(vs_vtk, self.faces[i])
self.gmm[i].replace_part(part_mesh)
else:
part_mesh = self.get_mesh_part(vs_vtk, base_mesh[1])
self.gmm[0].replace_part(part_mesh)
def set_brush(self, is_draw: bool):
self.render.set_brush(is_draw)
def replace_mesh(self, mesh: Optional[V_Mesh]):
mesh = torch.from_numpy(mesh[0]).float(), torch.from_numpy(mesh[1]).long()
self.add_mesh(mesh, for_slider=False)
# if mesh is None:
# return
# else:
# reduction = 1 - 50000. / mesh[1].shape[0]
# source_ = MeshStage.mesh_to_polydata(mesh)
# source_ = MeshStage.smooth_mesh(source_, ui_utils.SmoothingMethod.Taubin)
# self.decimate_mesh(source_, reduction, out=self.mapper.GetInput())
# self.is_changed = True
# if not self.to_init:
# self.to_init = True
# self.render.AddActor(self.actor)
def init_mesh_pos(self, vs: T):
vs = vs.clone()
r_a = rotation_utils.get_rotation_matrix(150, 1, degree=True)
r_b = rotation_utils.get_rotation_matrix(-15, 0, degree=True)
r = torch.from_numpy(np.einsum('km,mn->kn', r_b, r_a)).float()
vs = torch.einsum('ad,nd->na', r, vs)
vs[:, 0] += self.gmm_id * 2
return vs
@staticmethod
def mesh_to_polydata(mesh: Union[T_Mesh, V_Mesh], source: Optional[vtk.vtkPolyData] = None) -> vtk.vtkPolyData:
if source is None:
source = vtk.vtkPolyData()
vs, faces = mesh
if type(vs) is T:
vs, faces = vs.detach().cpu().numpy(), faces.detach().cpu().numpy()
new_vs_vtk = numpy_support.numpy_to_vtk(vs)
cells_npy = np.column_stack(
[np.full(faces.shape[0], 3, dtype=np.int64), faces.astype(np.int64)]).ravel()
vs_vtk, faces_vtk = vtk.vtkPoints(), vtk.vtkCellArray()
vs_vtk.SetData(new_vs_vtk)
faces_vtk.SetCells(faces.shape[0], numpy_support.numpy_to_vtkIdTypeArray(cells_npy))
source.SetPoints(vs_vtk)
source.SetPolys(faces_vtk)
return source
@property
def included(self):
for g in self.gmm:
if g.included:
return True
return False
def move_mesh_to_end(self, cycle: int):
self.offset += cycle
vs = None
for i in range(len(self)):
mapper = self.gmm[i].mapper
if mapper is not None and mapper.GetInput() is not None:
vs_vtk = mapper.GetInput().GetPoints()
if vs is None:
vs = numpy_support.vtk_to_numpy(vs_vtk.GetData())
vs[:, 0] += cycle * 2
vs_vtk.SetData(numpy_support.numpy_to_vtk(vs))
def pick(self, actor_address: str) -> bool:
return actor_address in self.addresses_dict
def __init__(self, opt: options.Options, shape_path: List[str], render: ui_utils.CanvasRender, render_number: int,
view_style: ui_utils.ViewStyle, to_init=True):
self.view_style = view_style
self.votes = {}
self.shape_id = shape_path[1]
self.gmm_id = render_number
self.render = render
self.symmetric_mode = sum(opt.symmetric) > 0 and False
self.selected = None
self.offset = render_number
# self.arrows = arrows.ArrowManger(render)
if self.shape_id != '-1':
self.base_mesh = files_utils.load_mesh( ''.join(shape_path))
self.raw_gmm = files_utils.load_gmm(f'{shape_path[0]}/{shape_path[1]}.txt', as_np=True)[:-1]
else:
self.base_mesh = None
self.raw_gmm = []
self.to_init = to_init
self.is_changed = False
self.gmm: List[gaussian_status.GaussianStatus] = self.add_gmm()
self.vs = self.faces = None
self.add_mesh(self.base_mesh)
self.addresses_dict: Dict[str, int] = {self.gmm[i].get_address(): i for i in range(len(self.gmm))}
if self.symmetric_mode:
for i in range(len(self) // 2):
self.make_twins(self.gmm[i].get_address(), self.gmm[i + len(self) // 2].get_address())
self.toggle_all()
# if self.raw_gmm:
# gmms = self.get_gmm()[0]
# files_utils.export_gmm(gmms, 0, f"./{render_number}")
class GmmStatuses:
def __len__(self):
return len(self.gmms)
def switch_arrows(self, arrow_type: ui_utils.Buttons):
self.main_gmm.switch_arrows(arrow_type)
def turn_gmm_off(self):
self.main_gmm.turn_gmm_off()
def turn_gmm_on(self):
self.main_gmm.turn_gmm_on()
def update_gmm(self, button: ui_utils.Buttons, key: str) -> bool:
return self.main_gmm.update_gmm(button, key)
def toggle_symmetric(self, force_include: bool = False):
for gmm in self.gmms:
gmm.toggle_symmetric(force_include)
def event_manger(self, object_id: str):
for gmm in self.gmms:
if gmm.event_manger(object_id):
return True
return False
def toggle_inclusion(self, object_id: str):
for gmm in self.gmms:
if gmm.toggle_inclusion(object_id)[0]:
return True
return False
@property
def main_gmm(self) -> GmmMeshStage:
return self.gmms[0]
def reset(self):
for gmm in self.gmms:
gmm.reset()
def set_brush(self, is_draw: bool):
for gmm in self.gmms:
gmm.set_brush(is_draw)
def move_mesh_to_end(self, ptr: int):
self.gmms[ptr].move_mesh_to_end(len(self))
def pick(self, actor_address: str) -> Optional[GmmMeshStage]:
for gmm in self.gmms:
if gmm.pick(actor_address):
return gmm
return None
def __init__(self, opt: options.Options, shape_paths: List[List[str]], render, view_styles: List[ui_utils.ViewStyle]):
self.gmms = [GmmMeshStage(opt, shape_path, render, i, view_style) for i, (shape_path, view_style) in
enumerate(zip(shape_paths, view_styles))]
def to_local(func):
def inner(self: MeshGmmStatuses.TransitionController, mouse_pos: Optional[Tuple[int, int]], *args, **kwargs):
if mouse_pos is not None:
size = self.render.GetRenderWindow().GetScreenSize()
aspect = self.render.GetAspect()
mouse_pos = float(mouse_pos[0]) / size[0] - .5, float(mouse_pos[1]) / size[1] - .5
mouse_pos = torch.tensor([mouse_pos[0] / aspect[1], mouse_pos[1] / aspect[0]])
return func(self, mouse_pos, *args, **kwargs)
return inner
class MeshGmmStatuses(GmmStatuses):
def aggregate_votes(self, select: bool):
if self.cur_canvas < len(self.gmms):
stage = self.gmms[self.cur_canvas]
changed = stage.aggregate_votes()
changed = list(filter(lambda x: not stage.gmm[x].disabled and stage.gmm[x].is_selected != select, changed))
for item in changed:
stage.gmm[item].toggle_selection()
return len(changed) > 0
def vote(self, *actors: Optional[vtk.vtkActor]):
self.gmms[self.cur_canvas].vote(*actors)
def init_draw(self, side: int):
self.cur_canvas = side
def sort_gmms(self, gmms, included):
order = torch.arange(gmms[0].shape[2]).tolist()
order = sorted(order, key=lambda x: included[x][0] * 100 + included[x][1])
gmms = [[item[:, :, order[i]] for item in gmms] for i in range(gmms[0].shape[2])]
gmms = [torch.stack([gmms[j][i] for j in range(len(gmms))], dim=2) for i in range(len(gmms[0]))]
return gmms
def save_light(self, root, gmms):
gmms = self.sort_gmms(*gmms)
save_dict = {'ids': {
gmm.shape_id: [gaussian.gaussian_id[1] for gaussian in gmm.gmm if gaussian.included]
for gmm in self.gmms if gmm.included},
'gmm': gmms}
path = f"{root}/{files_utils.get_time_name('light')}"
files_utils.save_pickle(save_dict, path)
def save(self, root: str, gmms):
# for gmm in self.gmms:
# if gmm.included:
# gmm.save(root)
if len(gmms[0]) > 0:
self.save_light(root, gmms)
def set_brush(self, is_draw: bool):
super(MeshGmmStatuses, self).set_brush(is_draw)
self.main_gmm.render.set_brush(is_draw)
def update_mesh(self, res=128):
if self.model_process is not None:
self.model_process.get_mesh(res)
return True
return False
# self.all_info[side] = gaussian_inds
def request_gmm(self) -> Tuple[TS, T]:
gmm, included = self.main_gmm.get_gmm()
return gmm, included
def replace_mesh(self):
if self.model_process is not None:
self.model_process.replace_mesh()
def exit(self):
if self.model_process is not None:
self.model_process.exit()
@property
def main_stage(self) -> GmmMeshStage:
return self.gmms[0]
@property
def stages(self):
return self.gmms
class TransitionController:
@property
def moving_axis(self) -> int:
return {ui_utils.EditDirection.X_Axis: 0,
ui_utils.EditDirection.Y_Axis: 2,
ui_utils.EditDirection.Z_Axis: 1}[self.edit_direction]
def get_delta_translation(self, mouse_pos: T) -> ARRAY:
delta_3d = np.zeros(3)
axis = self.moving_axis
vec = mouse_pos - self.origin_mouse
delta = torch.einsum('d,d', vec, self.dir_2d[:, axis])
delta_3d[axis] = delta
return delta_3d
def get_delta_rotation(self, mouse_pos: T) -> ARRAY:
projections = []
for pos in (self.origin_mouse, mouse_pos):
vec = pos - self.transition_origin_2d
projection = torch.einsum('d,da->a', vec, self.dir_2d)
projection[self.moving_axis] = 0
projection = nnf.normalize(projection, p=2, dim=0)
projections.append(projection)
sign = (projections[0][(self.moving_axis + 2) % 3] * projections[1][(self.moving_axis + 1) % 3]
- projections[0][(self.moving_axis + 1) % 3] * projections[1][(self.moving_axis + 2) % 3] ).sign()
angle = (torch.acos(torch.einsum('d,d', *projections)) * sign).item()
return ui_utils.get_rotation_matrix(angle, self.moving_axis)
def get_delta_scaling(self, mouse_pos: T) -> ARRAY:
raise NotImplementedError
def toggle_edit_direction(self, direction: ui_utils.EditDirection):
self.edit_direction = direction
@to_local
def get_transition(self, mouse_pos: Optional[T]) -> ui_utils.Transition:
transition = ui_utils.Transition(self.transition_origin.numpy(), self.transition_type)
if mouse_pos is not None:
if self.transition_type is ui_utils.EditType.Translating:
transition.translation = self.get_delta_translation(mouse_pos)
elif self.transition_type is ui_utils.EditType.Rotating:
transition.rotation = self.get_delta_rotation(mouse_pos)
elif self.transition_type is ui_utils.EditType.Scaling:
transition.rotation = self.get_delta_scaling(mouse_pos)
return transition
@to_local
def init_transition(self, mouse_pos: Tuple[int, int], transition_origin: T, transition_type: ui_utils.EditType):
transform_mat_vtk = self.camera.GetViewTransformMatrix()
dir_2d = torch.zeros(3, 4)
for i in range(3):
for j in range(4):
dir_2d[i, j] = transform_mat_vtk.GetElement(i, j)
self.transition_origin = transition_origin
transition_origin = torch.tensor(transition_origin.tolist() + [1])
transition_origin_2d = torch.einsum('ab,b->a', dir_2d, transition_origin)
self.transition_origin_2d = transition_origin_2d[:2] / transition_origin_2d[-1].abs()
# print(f"<{self.transition_origin[0]}, {self.transition_origin[1]}>")
# print(mouse_pos)
self.origin_mouse, self.dir_2d = mouse_pos, nnf.normalize(dir_2d[:2, :3], p=2, dim=1)
self.transition_type = transition_type
@property
def camera(self):
return self.render.GetActiveCamera()
def __init__(self, render: ui_utils.CanvasRender):
self.render = render
self.transition_origin = torch.zeros(3)
self.transition_origin_2d = torch.zeros(2)
self.origin_mouse, self.dir_2d = torch.zeros(2), torch.zeros(2, 3)
self.edit_direction = ui_utils.EditDirection.X_Axis
self.transition_type = ui_utils.EditType.Translating
@property
def selected_gaussians(self) -> Iterable[gaussian_status.GaussianStatus]:
return filter(lambda x: x.is_selected, self.main_stage.gmm)
def temporary_transition(self, mouse_pos: Optional[Tuple[int, int]] = None, end=False) -> bool:
transition = self.transition_controller.get_transition(mouse_pos)
is_change = False
for gaussian in self.selected_gaussians:
if end:
is_change = gaussian.end_transition(transition) or is_change
else:
is_change = gaussian.temporary_transition(transition) or is_change
return is_change
def end_transition(self, mouse_pos: Optional[Tuple[int, int]]) -> bool:
return self.temporary_transition(mouse_pos, True)
def init_transition(self, mouse_pos, transition_type: ui_utils.EditType):
center = list(map(lambda x: x.mu_baked, self.selected_gaussians))
if len(center) == 0:
return
# center = torch.from_numpy(np.stack(center, axis=0).mean(0))
center = torch.zeros(3)
self.transition_controller.init_transition(mouse_pos, center, transition_type)
def toggle_edit_direction(self, direction: ui_utils.EditDirection):
self.transition_controller.toggle_edit_direction(direction)
def clear_selection(self) -> bool:
is_changed = False
for gaussian in self.selected_gaussians:
gaussian.toggle_selection()
is_changed = True
return is_changed
def __init__(self, opt: options.Options, shape_paths: List[List[str]], render, view_styles: List[ui_utils.ViewStyle],
with_model: bool):
super(MeshGmmStatuses, self).__init__(opt, shape_paths, render, view_styles)
if with_model:
self.model_process = inference_processing.InferenceProcess(opt, self.main_stage.replace_mesh,
self.request_gmm,
shape_paths)
else:
self.model_process = None
self.counter = 0
self.cur_canvas = 0
self.transition_controller = MeshGmmStatuses.TransitionController(self.main_stage.render)
class MeshGmmUnited(MeshGmmStatuses):
def save(self, root: str):
super(MeshGmmUnited, self).save(root)
self.main_gmm.save(root, filter_by_selection)
def aggregate_votes(self, select: bool):
if self.cur_canvas < len(self.gmms):
stage = self.gmms[self.cur_canvas]
changed = stage.aggregate_votes()
changed = list(filter(lambda x: not stage.gmm[x].disabled and stage.gmm[x].included != select, changed))
for item in changed:
is_toggled, toggled = stage.toggle_inclusion_by_id(item, select)
if is_toggled:
if toggled[0].included:
new_addresses = self.main_gmm.add_gaussians(toggled)
for gaussian, new_address in zip(toggled, new_addresses):
self.stage_mapper[gaussian.get_address()] = new_address
self.make_twins(toggled, new_addresses)
else:
addresses = [gaussian.get_address() for gaussian in toggled]
addresses = list(filter(lambda x: x in self.stage_mapper, addresses))
self.main_gmm.remove_gaussians([self.stage_mapper[address] for address in addresses])
for address in addresses:
del self.stage_mapper[address]
return len(changed) > 0
else:
return self.update_selection(select)
def update_selection(self, select: bool):
changed = self.main_stage.aggregate_votes()
changed = filter(lambda x: self.main_stage.gmm[x].is_selected != select, changed)
for item in changed:
self.main_stage.gmm[item].toggle_selection()
return False
def vote(self, *actors: Optional[vtk.vtkActor]):
if self.cur_canvas < len(self.gmms):
self.gmms[self.cur_canvas].vote(*actors)
else:
self.main_gmm.vote(*actors)
def reset(self):
super(MeshGmmUnited, self).reset()
self.main_gmm.remove_all()
for gmm in self.gmms:
gmm.toggle_all()
self.stage_mapper = {}
def event_manger(self, object_id: str):
return self.toggle_inclusion(object_id) or self.main_gmm.event_manger(object_id)
def make_twins(self, toggled: List[gaussian_status.GaussianStatus], new_addresses : List[str]):
if len(new_addresses) == 2:
self.main_gmm.make_twins(*new_addresses)
else:
if toggled[0].twin is not None and toggled[0].twin.get_address() in self.stage_mapper:
self.main_gmm.make_twins(new_addresses[0], self.stage_mapper[toggled[0].twin.get_address()])
def toggle_symmetric(self, force_include: bool = False):
super(MeshGmmUnited, self).toggle_symmetric(force_include)
self.main_gmm.toggle_symmetric(force_include)
@property
def main_gmm(self) -> GmmMeshStage:
return self.main_gmm_
@property
def main_stage(self) -> GmmMeshStage:
return self.main_gmm_
def __init__(self, opt: options.Options, gmm_paths: List[int], renders_right, view_styles: List[ui_utils.ViewStyle],
main_render: ui_utils.CanvasRender, with_model: bool):
self.main_gmm_ = GmmMeshStage(opt, -1, main_render, len(gmm_paths), view_styles[-1], to_init=False)
super(MeshGmmUnited, self).__init__(opt, gmm_paths, renders_right, view_styles[:-1], with_model)
self.main_render = main_render
self.reset()
self.stage_mapper: Dict[str, str] = {}
def main():
opt = options.Options(tag="chairs_sym_hard").load()
model = train_utils.model_lc(opt)[0]
model = model.to(CPU)
colors = torch.rand(opt.num_gaussians, 3)
shape_nums = 1103, 1637, 2954, 3631, 4814
for shape_num in shape_nums:
mesh = files_utils.load_mesh(f"{opt.cp_folder}/occ/samples_{shape_num}")
gmm = files_utils.load_gmm(f"{opt.cp_folder}/gmms/samples_{shape_num}")
vs, faces = mesh
phi, mu, eigen, p, _ = [item.unsqueeze(0).unsqueeze(0) for item in gmm]
gmm = mu, p, phi, eigen
attention = model.get_attention(vs.unsqueeze(0), torch.tensor([shape_num], dtype=torch.int64))[-4:]
# _, supports = gm_utils.hierarchical_gm_log_likelihood_loss([gmm], vs.unsqueeze(0), get_supports=True)
# supports = supports[0][0]
supports = torch.cat(attention, dim=0)
supports = supports.mean(-1).mean(0)
label = supports.argmax(1)
colors_ = colors[label]
files_utils.export_mesh((vs, faces), f"{constants.OUT_ROOT}/{opt.tag}_{shape_num}b", colors=colors_)
return 0
if __name__ == '__main__':
from utils import train_utils
exit(main())
|