File size: 17,445 Bytes
55e58d1 |
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 |
import open3d as o3d
import numpy as np
import os
import cv2
from evaluation.constants import SCANNET_LABELS, SCANNET_IDS, SCANNET18_LABELS, SCANNET18_IDS, SCANNETPP84_IDS, SCANNETPP84_LABELS, SCANNET20_LABELS, SCANNET20_IDS, ARKIT_LABELS, ARKIT_IDS
class ScanNetDataset:
def __init__(self, seq_name, root='data/scannet', use_templates=False) -> None:
self.seq_name = seq_name
self.use_templates = use_templates
self.root = os.path.join(root, 'processed', seq_name)
self.rgb_dir = f'{self.root}/color'
self.depth_dir = f'{self.root}/depth'
self.segmentation_dir = f'{self.root}/output/mask'
self.object_dict_dir = f'{self.root}/output/object'
self.point_cloud_path = f'{self.root}/{seq_name}_vh_clean_2.ply'
self.mesh_path = self.point_cloud_path
self.extrinsics_dir = f'{self.root}/pose'
self.intrinsic_dir = f'{self.root}/intrinsic'
self.label_features_dict = None
self.depth_scale = 1000.0
self.image_size = self.get_image_size()
self.depth_size = self.get_depth_shape()
def get_frame_list(self, stride):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
end = int(image_list[-1].split('.')[0]) + 1
frame_id_list = [int(a.split('.')[0]) for a in image_list]
return list(frame_id_list)
def get_image_size(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
image_path = os.path.join(self.rgb_dir, image_list[0])
image = cv2.imread(image_path)
return image.shape[:2][::-1]
def get_depth_shape(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
depth_path = os.path.join(self.depth_dir, f"{image_list[0].split('.')[0]}.png")
depth = cv2.imread(depth_path, -1)
return depth.shape[:2][::-1]
def get_intrinsics(self, frame_id):
intrinsic_path = f'{self.intrinsic_dir}/intrinsic_depth.txt'
intrinsics = np.loadtxt(intrinsic_path)
intrinisc_cam_parameters = o3d.camera.PinholeCameraIntrinsic()
intrinisc_cam_parameters.set_intrinsics(self.image_size[0], self.image_size[1], intrinsics[0, 0], intrinsics[1, 1], intrinsics[0, 2], intrinsics[1, 2])
return intrinisc_cam_parameters
def get_extrinsic(self, frame_id):
pose_path = os.path.join(self.extrinsics_dir, str(frame_id) + '.txt')
pose = np.loadtxt(pose_path)
return pose
def get_depth(self, frame_id):
depth_path = os.path.join(self.depth_dir, str(frame_id) + '.png')
depth = cv2.imread(depth_path, -1)
depth = depth / self.depth_scale
depth = depth.astype(np.float32)
return depth
def get_rgb(self, frame_id, change_color=True):
rgb_path = os.path.join(self.rgb_dir, str(frame_id) + '.jpg')
rgb = cv2.imread(rgb_path)
if change_color:
rgb = cv2.cvtColor(rgb, cv2.COLOR_BGR2RGB)
return rgb
def get_segmentation(self, frame_id, align_with_depth=False):
segmentation_path = os.path.join(self.segmentation_dir, f'{frame_id}.png')
if not os.path.exists(segmentation_path):
assert False, f"Segmentation not found: {segmentation_path}"
segmentation = cv2.imread(segmentation_path, cv2.IMREAD_UNCHANGED)
if align_with_depth:
segmentation = cv2.resize(segmentation, self.depth_size, interpolation=cv2.INTER_NEAREST)
return segmentation
def get_frame_path(self, frame_id):
rgb_path = os.path.join(self.rgb_dir, str(frame_id) + '.jpg')
segmentation_path = os.path.join(self.segmentation_dir, f'{frame_id}.png')
return rgb_path, segmentation_path
def get_label_features(self):
if self.label_features_dict is None:
if self.use_templates:
label_features_dict = np.load(f'data/text_features/scannet_templates.npy', allow_pickle=True).item()
else:
label_features_dict = np.load(f'data/text_features/scannet.npy', allow_pickle=True).item()
self.label_features_dict = label_features_dict
return self.label_features_dict
def get_scene_points(self):
mesh = o3d.io.read_point_cloud(self.point_cloud_path)
vertices = np.asarray(mesh.points)
return vertices
def get_label_id(self):
self.class_id = SCANNET_IDS
self.class_label = SCANNET_LABELS
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label
class ARKitDataset(ScanNetDataset):
def __init__(self, seq_name, root='data/arkit_dust3r_posed'):
super().__init__(seq_name, root)
self.image_size = self.get_image_size()
def get_image_size(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
image_path = os.path.join(self.rgb_dir, image_list[0])
image = cv2.imread(image_path)
return image.shape[:2][::-1]
def get_frame_list(self, stride):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
end = int(image_list[-1].split('.')[0]) + 1
frame_id_list = [a.split('.')[0] for a in image_list]
return list(frame_id_list)
def get_label_id(self):
self.class_id = ARKIT_IDS
self.class_label = ARKIT_LABELS
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label
def get_label_features(self):
label_features_dict = np.load(f'data/text_features/arkit.npy', allow_pickle=True).item()
return label_features_dict
class ITWDataset(ARKitDataset):
def get_image_size(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('_')[0]))
image_path = os.path.join(self.rgb_dir, image_list[0])
image = cv2.imread(image_path)
return image.shape[:2][::-1]
def get_depth_shape(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('_')[0]))
depth_path = os.path.join(self.depth_dir, f"{image_list[0].split('.')[0]}.png")
depth = cv2.imread(depth_path, -1)
return depth.shape[:2][::-1]
def get_frame_list(self, stride):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('_')[0]))
frame_id_list = [a.split('.')[0] for a in image_list]
return list(frame_id_list)
def get_label_features(self):
label_features_dict = np.load(f'{self.root}/text_features.npy', allow_pickle=True).item()
return label_features_dict
def get_label_id(self):
text_features = self.get_label_features()
self.class_label = list(text_features.keys())
self.class_id = list(range(len(self.class_label)))
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label
class WildDataset(ARKitDataset):
def __init__(self, seq_name, root):
self.root = os.path.join(root, seq_name)
self.rgb_dir = f'{self.root}/images'
self.depth_dir = f'{self.root}/depth'
self.segmentation_dir = f'{self.root}/output/mask'
self.object_dict_dir = f'{self.root}/output/object'
self.point_cloud_path = f'{self.root}/point_cloud.ply'
self.mesh_path = self.point_cloud_path
self.extrinsics_dir = f'{self.root}/pose'
self.intrinsic_dir = f'{self.root}/intrinsic'
self.label_features_dict = None
self.depth_scale = 1000.0
self.image_size = self.get_depth_shape()
self.depth_size = self.get_depth_shape()
def get_label_features(self):
label_features_dict = np.load(f'{self.root}/text_features.npy', allow_pickle=True).item()
return label_features_dict
def get_segmentation(self, frame_id, align_with_depth=False):
segmentation_path = os.path.join(self.segmentation_dir, f'{frame_id}.png')
if not os.path.exists(segmentation_path):
assert False, f"Segmentation not found: {segmentation_path}"
segmentation = cv2.imread(segmentation_path, cv2.IMREAD_UNCHANGED)
segmentation = cv2.resize(segmentation, self.depth_size, interpolation=cv2.INTER_NEAREST)
return segmentation
def get_label_id(self):
text_features = self.get_label_features()
self.class_label = list(text_features.keys())
self.class_id = list(range(len(self.class_label)))
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label
class ScannetPP2Dataset(ScanNetDataset):
def __init__(self, seq_name, root='data/scannetpp_dust3r_posed'):
super().__init__(seq_name, root)
self.image_size = self.get_image_size()
self.depth_size = self.get_depth_shape()
self.point_cloud_path = f'{self.root}/{seq_name}.ply'
def get_image_size(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0].split('_')[1]))
image_path = os.path.join(self.rgb_dir, image_list[0])
image = cv2.imread(image_path)
return image.shape[:2][::-1]
def get_depth_shape(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0].split('_')[1]))
depth_path = os.path.join(self.depth_dir, f"{image_list[0].split('.')[0]}.png")
depth = cv2.imread(depth_path, -1)
return depth.shape[:2][::-1]
def get_frame_list(self, stride):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0].split('_')[1]))
frame_id_list = [a.split('.')[0] for a in image_list]
return list(frame_id_list)
def get_segmentation(self, frame_id, align_with_depth=False):
segmentation_path = os.path.join(self.segmentation_dir, f'{frame_id}.png')
if not os.path.exists(segmentation_path):
assert False, f"Segmentation not found: {segmentation_path}"
segmentation = cv2.imread(segmentation_path, cv2.IMREAD_UNCHANGED)
segmentation = cv2.resize(segmentation, self.depth_size, interpolation=cv2.INTER_NEAREST)
return segmentation
def get_label_id(self):
self.class_id = SCANNETPP84_IDS
self.class_label = SCANNETPP84_LABELS
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label
def get_label_features(self):
label_features_dict = np.load(f'data/text_features/scannetpp84.npy', allow_pickle=True).item()
return label_features_dict
def get_depth(self, frame_id):
depth_path = os.path.join(self.depth_dir, str(frame_id) + '.png')
depth = cv2.imread(depth_path, -1)
depth = depth / self.depth_scale
depth = depth.astype(np.float32)
return depth
def get_intrinsics(self, frame_id):
intrinsic_path = f'{self.intrinsic_dir}/intrinsic_depth.txt'
intrinsics = np.loadtxt(intrinsic_path)
intrinisc_cam_parameters = o3d.camera.PinholeCameraIntrinsic()
intrinisc_cam_parameters.set_intrinsics(self.image_size[0], self.image_size[1], intrinsics[0, 0], intrinsics[1, 1], intrinsics[0, 2], intrinsics[1, 2])
return intrinisc_cam_parameters
class ScanNet18Dataset:
def __init__(self, seq_name, root='data/scannet') -> None:
self.seq_name = seq_name
self.root = os.path.join(root, 'processed', seq_name)
self.rgb_dir = f'{self.root}/color'
self.depth_dir = f'{self.root}/depth'
self.segmentation_dir = f'{self.root}/output/mask'
self.object_dict_dir = f'{self.root}/output/object'
self.point_cloud_path = f'{self.root}/{seq_name}.ply'
self.mesh_path = self.point_cloud_path
self.extrinsics_dir = f'{self.root}/pose'
self.intrinsic_dir = f'{self.root}/intrinsic'
self.depth_scale = 1000.0
self.image_size = self.get_image_size()
self.depth_size = self.get_depth_shape()
def get_frame_list(self, stride):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
end = int(image_list[-1].split('.')[0]) + 1
frame_id_list = [a.split('.')[0] for a in image_list]
return list(frame_id_list)
def get_image_size(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
image_path = os.path.join(self.rgb_dir, image_list[0])
image = cv2.imread(image_path)
return image.shape[:2][::-1]
def get_depth_shape(self):
image_list = os.listdir(self.rgb_dir)
image_list = sorted(image_list, key=lambda x: int(x.split('.')[0]))
depth_path = os.path.join(self.depth_dir, f"{image_list[0].split('.')[0]}.png")
depth = cv2.imread(depth_path, -1)
return depth.shape[:2][::-1]
def get_intrinsics(self, frame_id):
intrinsic_path = f'{self.intrinsic_dir}/intrinsic_depth.txt'
intrinsics = np.loadtxt(intrinsic_path)
intrinisc_cam_parameters = o3d.camera.PinholeCameraIntrinsic()
intrinisc_cam_parameters.set_intrinsics(self.image_size[0], self.image_size[1], intrinsics[0, 0], intrinsics[1, 1], intrinsics[0, 2], intrinsics[1, 2])
return intrinisc_cam_parameters
def get_extrinsic(self, frame_id):
pose_path = os.path.join(self.extrinsics_dir, str(frame_id) + '.txt')
pose = np.loadtxt(pose_path)
return pose
def get_depth(self, frame_id):
depth_path = os.path.join(self.depth_dir, str(frame_id) + '.png')
depth = cv2.imread(depth_path, -1)
depth = depth / self.depth_scale
depth = depth.astype(np.float32)
return depth
def get_rgb(self, frame_id, change_color=True):
rgb_path = os.path.join(self.rgb_dir, str(frame_id) + '.jpg')
rgb = cv2.imread(rgb_path)
if change_color:
rgb = cv2.cvtColor(rgb, cv2.COLOR_BGR2RGB)
return rgb
def get_segmentation(self, frame_id, align_with_depth=False):
segmentation_path = os.path.join(self.segmentation_dir, f'{frame_id}.png')
if not os.path.exists(segmentation_path):
assert False, f"Segmentation not found: {segmentation_path}"
segmentation = cv2.imread(segmentation_path, cv2.IMREAD_UNCHANGED)
segmentation = cv2.resize(segmentation, self.depth_size, interpolation=cv2.INTER_NEAREST)
return segmentation
def get_frame_path(self, frame_id):
rgb_path = os.path.join(self.rgb_dir, str(frame_id) + '.jpg')
segmentation_path = os.path.join(self.segmentation_dir, f'{frame_id}.png')
return rgb_path, segmentation_path
def get_label_features(self):
label_features_dict = np.load(f'data/text_features/scannet18.npy', allow_pickle=True).item()
return label_features_dict
def get_scene_points(self):
mesh = o3d.io.read_point_cloud(self.point_cloud_path)
vertices = np.asarray(mesh.points)
return vertices
def get_label_id(self):
self.class_id = SCANNET18_IDS
self.class_label = SCANNET18_LABELS
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label
class ScanNet20Dataset(ScanNet18Dataset):
def __init__(self, *args, **kwargs) -> None:
super().__init__(*args, **kwargs)
self.point_cloud_path = f'{self.root}/{self.seq_name}_vh_clean_2.ply'
def get_label_features(self):
label_features_dict = np.load(f'/home/jovyan/users/lemeshko/Indoor/MaskClustering/data/text_features/scannet20.npy', allow_pickle=True).item()
return label_features_dict
def get_label_id(self):
self.class_id = SCANNET20_IDS
self.class_label = SCANNET20_LABELS
self.label2id = {}
self.id2label = {}
for label, id in zip(self.class_label, self.class_id):
self.label2id[label] = id
self.id2label[id] = label
return self.label2id, self.id2label |