Spaces:
Running
on
Zero
Running
on
Zero
File size: 21,094 Bytes
05fb4ab |
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 |
import torch
import torch.nn.functional as F
import torch.utils.data
from packaging import version
def pre_process_image_glunet(source_img, device, mean_vector=[0.485, 0.456, 0.406],
std_vector=[0.229, 0.224, 0.225]):
"""
Image is in range [0, 255}. Creates image at 256x256, and applies imagenet weights to both.
Args:
source_img: torch tensor, bx3xHxW in range [0, 255], not normalized yet
device:
mean_vector:
std_vector:
Returns:
image at original and 256x256 resolution
"""
# img has shape bx3xhxw
b, _, h_scale, w_scale = source_img.shape
source_img_copy = source_img.float().to(device).div(255.0)
# mean = torch.as_tensor(mean_vector, dtype=source_img_copy.dtype, device=source_img_copy.device)
# std = torch.as_tensor(std_vector, dtype=source_img_copy.dtype, device=source_img_copy.device)
# source_img_copy.sub_(mean[:, None, None]).div_(std[:, None, None])
# resolution 256x256
source_img_256 = torch.nn.functional.interpolate(input=source_img.float().to(device).div(255.0), size=(256, 256), mode='area')
# source_img_256 = source_img_256.float().div(255.0)
# source_img_256.sub_(mean[:, None, None]).div_(std[:, None, None])
return source_img_copy.to(device), source_img_256.to(device)
class CATsBatchPreprocessing:
""" Class responsible for processing the mini-batch to create the desired training inputs for GLU-Net based networks.
Particularly, from the source and target images at original resolution as well as the corresponding ground-truth
flow field, needs to create the source, target and flow at resolution 256x256 for training the L-Net.
"""
def __init__(self, settings, apply_mask=False, apply_mask_zero_borders=False, sparse_ground_truth=False,
mapping=False):
"""
Args:
settings: settings
apply_mask: apply ground-truth correspondence mask for loss computation?
apply_mask_zero_borders: apply mask zero borders (equal to 0 at black borders in target image) for loss
computation?
sparse_ground_truth: is ground-truth sparse? Important for downscaling/upscaling of the flow field
mapping: load correspondence map instead of flow field?
"""
self.apply_mask = apply_mask
self.apply_mask_zero_borders = apply_mask_zero_borders
self.sparse_ground_truth = sparse_ground_truth
self.device = getattr(settings, 'device', None)
if self.device is None:
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
self.mapping = mapping
def __call__(self, mini_batch, *args, **kwargs):
"""
args:
mini_batch: The input data, should contain the following fields:
'source_image', 'target_image', 'correspondence_mask'
'flow_map' if self.mapping is False else 'correspondence_map_pyro'
'mask_zero_borders' if self.apply_mask_zero_borders
returns:
TensorDict: output data block with following fields:
'source_image', 'target_image', 'source_image_256', 'target_image_256', flow_map',
'flow_map_256', 'mask', 'mask_256', 'correspondence_mask'
"""
source_image, source_image_256 = pre_process_image_glunet(mini_batch['source_image'], self.device)
target_image, target_image_256 = pre_process_image_glunet(mini_batch['target_image'], self.device)
# At original resolution
if self.sparse_ground_truth:
flow_gt_original = mini_batch['flow_map'][0].to(self.device)
flow_gt_256 = mini_batch['flow_map'][1].to(self.device)
if flow_gt_original.shape[1] != 2:
# shape is bxhxwx2
flow_gt_original = flow_gt_original.permute(0, 3, 1, 2)
if flow_gt_256.shape[1] != 2:
# shape is bxhxwx2
flow_gt_256 = flow_gt_256.permute(0, 3, 1, 2)
else:
if self.mapping:
mapping_gt_original = mini_batch['correspondence_map_pyro'].to(self.device)
# flow_gt_original = unormalise_and_convert_mapping_to_flow(mapping_gt_original.permute(0,3,1,2))
else:
flow_gt_original = mini_batch['flow_map'].to(self.device)
if flow_gt_original.shape[1] != 2:
# shape is bxhxwx2
flow_gt_original = flow_gt_original.permute(0, 3, 1, 2)
bs, _, h_original, w_original = flow_gt_original.shape
# now we have flow everywhere, at 256x256 resolution, b, _, 256, 256
flow_gt_256 = F.interpolate(flow_gt_original, (256, 256), mode='bilinear', align_corners=False)
flow_gt_256[:, 0, :, :] *= 256.0/float(w_original)
flow_gt_256[:, 1, :, :] *= 256.0/float(h_original)
bs, _, h_original, w_original = flow_gt_original.shape
bs, _, h_256, w_256 = flow_gt_256.shape
# defines the mask to use during training
mask = None
mask_256 = None
if self.apply_mask_zero_borders:
# make mask to remove all black borders
mask = mini_batch['mask_zero_borders'].to(self.device) # bxhxw, torch.uint8
mask_256 = F.interpolate(mask.unsqueeze(1).float(), (256, 256), mode='bilinear',
align_corners=False).squeeze(1).byte() # bx256x256, rounding
mask_256 = mask_256.bool() if version.parse(torch.__version__) >= version.parse("1.1") else mask_256.byte()
elif self.apply_mask:
if self.sparse_ground_truth:
mask = mini_batch['correspondence_mask'][0].to(self.device)
mask_256 = mini_batch['correspondence_mask'][1].to(self.device)
else:
mask = mini_batch['correspondence_mask'].to(self.device) # bxhxw, torch.uint8
mask_256 = F.interpolate(mask.unsqueeze(1).float(), (256, 256), mode='bilinear',
align_corners=False).squeeze(1).byte() # bx256x256, rounding
mask_256 = mask_256.bool() if version.parse(torch.__version__) >= version.parse("1.1") else mask_256.byte()
mini_batch['source_image'] = source_image
mini_batch['target_image'] = target_image
mini_batch['source_image_256'] = source_image_256
mini_batch['target_image_256'] = target_image_256
mini_batch['flow_map'] = flow_gt_original
mini_batch['flow_map_256'] = flow_gt_256
mini_batch['mask'] = mask
mini_batch['mask_256'] = mask_256
if self.sparse_ground_truth:
mini_batch['correspondence_mask'][0] = mini_batch['correspondence_mask'][0].to(self.device)
mini_batch['correspondence_mask'][1] = mini_batch['correspondence_mask'][1].to(self.device)
else:
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'].to(self.device)
return mini_batch
class GLUNetBatchPreprocessing:
""" Class responsible for processing the mini-batch to create the desired training inputs for GLU-Net based networks.
Particularly, from the source and target images at original resolution as well as the corresponding ground-truth
flow field, needs to create the source, target and flow at resolution 256x256 for training the L-Net.
"""
def __init__(self, settings, apply_mask=False, apply_mask_zero_borders=False, sparse_ground_truth=False,
mapping=False, megadepth=False):
"""
Args:
settings: settings
apply_mask: apply ground-truth correspondence mask for loss computation?
apply_mask_zero_borders: apply mask zero borders (equal to 0 at black borders in target image) for loss
computation?
sparse_ground_truth: is ground-truth sparse? Important for downscaling/upscaling of the flow field
mapping: load correspondence map instead of flow field?
"""
self.apply_mask = apply_mask
self.apply_mask_zero_borders = apply_mask_zero_borders
self.sparse_ground_truth = sparse_ground_truth
self.megadepth = megadepth
self.device = getattr(settings, 'device', None)
# if self.device is None:
# self.device = torch.device("cuda" if torch.cuda.is_available() and settings.use_gpu else "cpu")
print("Using device: {}".format(self.device))
self.mapping = mapping
def __call__(self, mini_batch, *args, **kwargs):
"""
args:
mini_batch: The input data, should contain the following fields:
'source_image', 'target_image', 'correspondence_mask'
'flow_map' if self.mapping is False else 'correspondence_map_pyro'
'mask_zero_borders' if self.apply_mask_zero_borders
returns:
TensorDict: output data block with following fields:
'source_image', 'target_image', 'source_image_256', 'target_image_256', flow_map',
'flow_map_256', 'mask', 'mask_256', 'correspondence_mask'
"""
source_image, source_image_256 = pre_process_image_glunet(mini_batch['source_image'], self.device)
target_image, target_image_256 = pre_process_image_glunet(mini_batch['target_image'], self.device)
# At original resolution
if self.sparse_ground_truth:
flow_gt_original = mini_batch['flow_map'][0].to(self.device)
flow_gt_256 = mini_batch['flow_map'][1].to(self.device)
if flow_gt_original.shape[1] != 2:
# shape is bxhxwx2
flow_gt_original = flow_gt_original.permute(0, 3, 1, 2)
if flow_gt_256.shape[1] != 2:
# shape is bxhxwx2
flow_gt_256 = flow_gt_256.permute(0, 3, 1, 2)
else:
if self.mapping:
mapping_gt_original = mini_batch['correspondence_map_pyro'][-1].to(self.device)
# flow_gt_original = unormalise_and_convert_mapping_to_flow(mapping_gt_original.permute(0,3,1,2))
else:
if self.megadepth:
flow_gt_original = mini_batch['flow_map'][0].to(self.device)
else:
flow_gt_original = mini_batch['flow_map'].to(self.device)
src_vis = mini_batch['source_image']
trg_vis = mini_batch['target_image']
flow_vis = mini_batch['flow_map'][0]
if flow_gt_original.shape[1] != 2:
# shape is bxhxwx2
flow_gt_original = flow_gt_original.permute(0, 3, 1, 2)
bs, _, h_original, w_original = flow_gt_original.shape
# now we have flow everywhere, at 256x256 resolution, b, _, 256, 256
flow_gt_256 = F.interpolate(flow_gt_original, (256, 256), mode='bilinear', align_corners=False)
flow_gt_256[:, 0, :, :] *= 256.0/float(w_original)
flow_gt_256[:, 1, :, :] *= 256.0/float(h_original)
bs, _, h_original, w_original = flow_gt_original.shape
bs, _, h_256, w_256 = flow_gt_256.shape
# defines the mask to use during training
mask = None
mask_256 = None
if self.apply_mask_zero_borders:
# make mask to remove all black borders
mask = mini_batch['mask_zero_borders'].to(self.device) # bxhxw, torch.uint8
mask_256 = F.interpolate(mask.unsqueeze(1).float(), (256, 256), mode='bilinear',
align_corners=False).squeeze(1).byte() # bx256x256, rounding
mask_256 = mask_256.bool() if version.parse(torch.__version__) >= version.parse("1.1") else mask_256.byte()
elif self.apply_mask:
if self.sparse_ground_truth:
mask = mini_batch['correspondence_mask'][0].to(self.device)
mask_256 = mini_batch['correspondence_mask'][1].to(self.device)
else:
if self.megadepth:
mask = mini_batch['correspondence_mask'][0].to(self.device) # bxhxw, torch.uint8
else:
mask = mini_batch['correspondence_mask'].to(self.device) # bxhxw, torch.uint8
mask_256 = F.interpolate(mask.unsqueeze(1).float(), (256, 256), mode='bilinear',
align_corners=False).squeeze(1).byte() # bx256x256, rounding
mask_256 = mask_256.bool() if version.parse(torch.__version__) >= version.parse("1.1") else mask_256.byte()
mini_batch['source_image'] = source_image
mini_batch['target_image'] = target_image
mini_batch['source_image_256'] = source_image_256
mini_batch['target_image_256'] = target_image_256
mini_batch['flow_map'] = flow_gt_original
mini_batch['flow_map_256'] = flow_gt_256
mini_batch['mask'] = mask
mini_batch['mask_256'] = mask_256
if self.sparse_ground_truth:
if self.megadepth:
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'][0].to(self.device)
else:
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'].to(self.device)
else:
if self.megadepth:
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'][0].to(self.device)
else:
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'].to(self.device)
return mini_batch
class DocBatchPreprocessing:
""" Class responsible for processing the mini-batch to create the desired training inputs for GLU-Net based networks.
Particularly, from the source and target images at original resolution as well as the corresponding ground-truth
flow field, needs to create the source, target and flow at resolution 256x256 for training the L-Net.
"""
def __init__(self, settings, apply_mask=False, apply_mask_zero_borders=False, sparse_ground_truth=False,
mapping=False, megadepth=False):
"""
Args:
settings: settings
apply_mask: apply ground-truth correspondence mask for loss computation?
apply_mask_zero_borders: apply mask zero borders (equal to 0 at black borders in target image) for loss
computation?
sparse_ground_truth: is ground-truth sparse? Important for downscaling/upscaling of the flow field
mapping: load correspondence map instead of flow field?
"""
self.apply_mask = apply_mask
self.apply_mask_zero_borders = apply_mask_zero_borders
self.sparse_ground_truth = sparse_ground_truth
self.megadepth = megadepth
self.device = getattr(settings, 'device', None)
# if self.device is None:
# self.device = torch.device("cuda" if torch.cuda.is_available() and settings.use_gpu else "cpu")
print("Using device: {}".format(self.device))
self.mapping = mapping
def __call__(self, mini_batch, *args, **kwargs):
"""
args:
mini_batch: The input data, should contain the following fields:
'source_image', 'target_image', 'correspondence_mask'
'flow_map' if self.mapping is False else 'correspondence_map_pyro'
'mask_zero_borders' if self.apply_mask_zero_borders
returns:
TensorDict: output data block with following fields:
'source_image', 'target_image', 'source_image_256', 'target_image_256', flow_map',
'flow_map_256', 'mask', 'mask_256', 'correspondence_mask'
"""
source_image, source_image_256 = pre_process_image_glunet(mini_batch['source_image'], self.device) # [24, 3, 512, 512],[24, 3, 256, 256])
# target_image, target_image_256 = pre_process_image_glunet(mini_batch['target_image'], self.device)
target_image = None
# At original resolution
if self.sparse_ground_truth: # false
flow_gt_original = mini_batch['flow_map'][0].to(self.device)
flow_gt_256 = mini_batch['flow_map'][1].to(self.device)
if flow_gt_original.shape[1] != 2:
# shape is bxhxwx2
flow_gt_original = flow_gt_original.permute(0, 3, 1, 2)
if flow_gt_256.shape[1] != 2:
# shape is bxhxwx2
flow_gt_256 = flow_gt_256.permute(0, 3, 1, 2)
else: # true
if self.mapping: # false
mapping_gt_original = mini_batch['correspondence_map_pyro'][-1].to(self.device)
# flow_gt_original = unormalise_and_convert_mapping_to_flow(mapping_gt_original.permute(0,3,1,2))
else: # true
if self.megadepth: # false
flow_gt_original = mini_batch['flow_map'][0].to(self.device)
elif 'flow_map' in mini_batch:# true
flow_gt_original = mini_batch['flow_map'].to(self.device) # [24, 2, 512, 512]
else:
mini_batch['source_image'] = source_image
mini_batch['source_image_256'] = source_image_256
return mini_batch
# src_vis = mini_batch['source_image']
# trg_vis = mini_batch['target_image']
# flow_vis = mini_batch['flow_map'][0]
if flow_gt_original.shape[1] != 2:
# shape is bxhxwx2
flow_gt_original = flow_gt_original.permute(0, 3, 1, 2)
bs, _, h_original, w_original = flow_gt_original.shape
# now we have flow everywhere, at 256x256 resolution, b, _, 256, 256
flow_gt_256 = F.interpolate(flow_gt_original, (256, 256), mode='bilinear', align_corners=False)
flow_gt_256[:, 0, :, :] *= 256.0/float(w_original)
flow_gt_256[:, 1, :, :] *= 256.0/float(h_original)
bs, _, h_original, w_original = flow_gt_original.shape
bs, _, h_256, w_256 = flow_gt_256.shape
# defines the mask to use during training
mask = None
mask_256 = None
if self.apply_mask_zero_borders: # false
# make mask to remove all black borders
mask = mini_batch['mask_zero_borders'].to(self.device) # bxhxw, torch.uint8
mask_256 = F.interpolate(mask.unsqueeze(1).float(), (256, 256), mode='bilinear',
align_corners=False).squeeze(1).byte() # bx256x256, rounding
mask_256 = mask_256.bool() if version.parse(torch.__version__) >= version.parse("1.1") else mask_256.byte()
elif self.apply_mask: # false
if self.sparse_ground_truth:
mask = mini_batch['correspondence_mask'][0].to(self.device)
mask_256 = mini_batch['correspondence_mask'][1].to(self.device)
else:
if self.megadepth:
mask = mini_batch['correspondence_mask'][0].to(self.device) # bxhxw, torch.uint8
else:
mask = mini_batch['correspondence_mask'].to(self.device) # bxhxw, torch.uint8
mask_256 = F.interpolate(mask.unsqueeze(1).float(), (256, 256), mode='bilinear',
align_corners=False).squeeze(1).byte() # bx256x256, rounding
mask_256 = mask_256.bool() if version.parse(torch.__version__) >= version.parse("1.1") else mask_256.byte()
mini_batch['source_image'] = source_image
if 'target_image' in mini_batch:
mini_batch['target_image'] = target_image
else:
mini_batch['target_image'] = None
mini_batch['source_image_256'] = source_image_256
mini_batch['target_image_256'] = None
mini_batch['flow_map'] = flow_gt_original
mini_batch['flow_map_256'] = flow_gt_256
mini_batch['mask'] = mask
mini_batch['mask_256'] = mask_256
if self.sparse_ground_truth: # false
if self.megadepth: # false
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'][0].to(self.device)
else:
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'].to(self.device)
else: # true
if self.megadepth: # false
mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'][0].to(self.device)
else: # true
pass
# mini_batch['correspondence_mask'] = mini_batch['correspondence_mask'].to(self.device)
return mini_batch
|