-
Notifications
You must be signed in to change notification settings - Fork 2
/
process.py
668 lines (545 loc) · 23.3 KB
/
process.py
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
from typing import Dict
import SimpleITK
import tqdm
import json
from pathlib import Path
import tifffile
import numpy as np
import pytorch_lightning as pl
from pl_bolts.models.autoencoders.components import (resnet18_decoder,
resnet18_encoder,
)
from pl_bolts.models.autoencoders import VAE
import torch
import pandas as pd
import numpy as np
from PIL import Image
import cv2
from pretrainedmodels import se_resnext50_32x4d
from efficientnet_pytorch import EfficientNet
import torchvision.models as models
import torch.nn as nn
from torchvision import transforms
from evalutils import ClassificationAlgorithm
from evalutils.validators import (
UniquePathIndicesValidator,
UniqueImagesValidator,
)
from evalutils.io import ImageLoader
GPU = torch.cuda.is_available()
if GPU:
device = "cuda"
else:
device = "cpu"
class VAE2(pl.LightningModule):
"""
Standard VAE with Gaussian Prior and approx posterior.
Model is available pretrained on different datasets:
"""
def __init__(
self,
input_height: int,
enc_type: str = 'resnet18',
first_conv: bool = False,
maxpool1: bool = False,
enc_out_dim: int = 512,
kl_coeff: float = 0.1,
latent_dim: int = 256,
lr: float = 1e-5,
**kwargs
):
"""
Args:
input_height: height of the images
enc_type: option between resnet18 or resnet50
first_conv: use standard kernel_size 7, stride 2 at start or
replace it with kernel_size 3, stride 1 conv
maxpool1: use standard maxpool to reduce spatial dim of feat by a factor of 2
enc_out_dim: set according to the out_channel count of
encoder used (512 for resnet18, 2048 for resnet50)
kl_coeff: coefficient for kl term of the loss
latent_dim: dim of latent space
lr: learning rate for Adam
"""
super(VAE2, self).__init__()
self.save_hyperparameters()
self.lr = lr
self.kl_coeff = kl_coeff
self.enc_out_dim = enc_out_dim
self.latent_dim = latent_dim
self.input_height = input_height
valid_encoders = {
'resnet18': {
'enc': resnet18_encoder,
'dec': resnet18_decoder,
},
}
self.encoder = valid_encoders[enc_type]['enc'](first_conv, maxpool1)
self.decoder = valid_encoders[enc_type]['dec'](self.latent_dim, self.input_height, first_conv, maxpool1)
self.fc_mu = nn.Linear(self.enc_out_dim, self.latent_dim)
self.fc_var = nn.Linear(self.enc_out_dim, self.latent_dim)
self.train_loss = 0
self.val_loss = 0
self.epoch = 0
self.images = []
self.input_images = []
self.val_step = 0
self.train_losses = []
def forward(self, x):
x = self.encoder(x)
mu = self.fc_mu(x)
log_var = self.fc_var(x)
p, q, z = self.sample(mu, log_var)
return self.decoder(z)
def _run_step(self, x):
x = self.encoder(x)
mu = self.fc_mu(x)
log_var = self.fc_var(x)
p, q, z = self.sample(mu, log_var)
return z, self.decoder(z), p, q
def sample(self, mu, log_var):
std = torch.exp(log_var / 2)
p = torch.distributions.Normal(torch.zeros_like(mu), torch.ones_like(std))
q = torch.distributions.Normal(mu, std)
z = q.rsample()
return p, q, z
def stepfadfadsf(self, batch, batch_idx):
x = batch
z, x_hat, p, q = self._run_step(x)
recon_loss = F.mse_loss(x_hat, x, reduction='mean')
log_qz = q.log_prob(z)
log_pz = p.log_prob(z)
kl = log_qz - log_pz
kl = kl.mean()
kl *= self.kl_coeff
loss = kl + recon_loss
logs = {
"recon_loss": recon_loss,
"kl": kl,
"loss": loss,
}
return loss, logs,
def training_step(self, batch, batch_idx):
x = batch
z, x_hat, p, q = self._run_step(x)
recon_loss = F.mse_loss(x_hat, x, reduction='mean')
log_qz = q.log_prob(z)
log_pz = p.log_prob(z)
kl = log_qz - log_pz
kl = kl.mean()
kl *= self.kl_coeff
loss = kl + recon_loss
#self.log_dict({f"train_{k}": v for k, v in logs.items()}, on_step=True, on_epoch=False)
self.train_losses.append(recon_loss.cpu().detach().numpy())
self.train_loss = recon_loss.cpu().detach().numpy()
return loss
def validation_step(self, batch, batch_idx):
x = batch
z, x_hat, p, q = self._run_step(x)
recon_loss = F.mse_loss(x_hat, x, reduction='mean')
log_qz = q.log_prob(z)
log_pz = p.log_prob(z)
kl = log_qz - log_pz
kl = kl.mean()
kl *= self.kl_coeff
loss = kl + recon_loss
#self.log_dict({f"val_{k}": v for k, v in logs.items()})
self.val_loss += recon_loss.cpu().detach().numpy()
imgs = x_hat.cpu().detach().numpy()
self.images.append([Image.fromarray((image*255).astype(np.uint8).transpose(1,2,0)) for image in imgs])
if self.epoch == 0 or self.epoch==1:
self.input_images.append([Image.fromarray((image*255).astype(np.uint8).transpose(1,2,0)) for image in x.cpu().detach().numpy()])
self.val_step += 1
return loss
def configure_optimizers(self):
return torch.optim.Adam(self.parameters(), lr=self.lr)
def validation_epoch_end(self, outputs):
#print(self.epoch)
self.val_loss = self.val_loss / self.val_step
if len(self.train_losses) > 626:
self.train_loss = np.mean(self.train_losses[:-625])
if self.epoch == 0 or self.epoch == 1:
self.images = [item for sublist in self.images for item in sublist]
self.input_images = [item for sublist in self.input_images for item in sublist]
logs = {
"train_loss": self.train_loss,
"val_loss": self.val_loss,
"epoch": self.epoch,
"recons": [wandb.Image(image) for image in self.images[:100]],
"originals": [wandb.Image(image) for image in self.input_images[:100]],
}
self.images = []
self.input_images = []
else:
self.images = [item for sublist in self.images for item in sublist]
logs = {
"train_loss": self.train_loss,
"val_loss": self.val_loss,
"epoch": self.epoch,
"recons": [wandb.Image(image) for image in self.images[:100]],
}
self.images = []
save_str = "VAE_resnet18" + "_epoch+15_" + str(self.epoch) + ".pth"
torch.save(self.state_dict(), save_str)
shutil.copy(save_str, os.path.join(wandb.run.dir, save_str))
wandb.save(os.path.join(wandb.run.dir, save_str))
wandb.log(logs)
self.epoch += 1
self.val_step = 0
class Autoencoder(nn.Module):
def __init__(self):
super(Autoencoder, self).__init__()
self.encoder = nn.Sequential( # like the Composition layer you built
nn.Conv2d(3, 8, 3, stride=2, padding=1), #128x128x8
nn.ReLU(),
nn.Conv2d(8, 16, 3, stride=2, padding=1), #64x64x16
nn.ReLU(),
nn.Conv2d(16, 32, 3, stride=2, padding=1),#32x32x32
nn.ReLU(),
nn.Conv2d(32, 64, 3, stride=2, padding=1), #16x16x64
nn.ReLU(),
nn.Conv2d(64, 128, 3, stride=2, padding=1), #8x8x128
nn.ReLU(),
nn.Conv2d(128, 256, 3, stride=2, padding=1), #4x4x256
nn.ReLU(),
nn.Conv2d(256, 512, 3, stride=2, padding=1), #4x4x512
)
self.decoder = nn.Sequential(
nn.ConvTranspose2d(512, 256, 3, stride=2, padding=1, output_padding=1),
nn.ReLU(),
nn.ConvTranspose2d(256, 128, 3, stride=2, padding=1, output_padding=1),
nn.ReLU(),
nn.ConvTranspose2d(128, 64, 3, stride=2, padding=1, output_padding=1),
nn.ReLU(),
nn.ConvTranspose2d(64, 32, 3, stride=2, padding=1, output_padding=1),
nn.ReLU(),
nn.ConvTranspose2d(32, 16, 3, stride=2, padding=1, output_padding=1),
nn.ReLU(),
nn.ConvTranspose2d(16, 8, 3, stride=2, padding=1, output_padding=1),
nn.ReLU(),
nn.ConvTranspose2d(8, 3, 3, stride=2, padding=1, output_padding=1),
nn.Sigmoid()
)
def forward(self, x):
x = self.encoder(x)
x = self.decoder(x)
return x
def getOD(model, image):
results = model(image, size=288).pandas().xywhn[0]
cols = ["xcenter", "ycenter", "width", "height", "confidence", "class", "name"]
df = pd.DataFrame([[0.5, 0.5, 1, 1, 0.001, 0, "OD"]], columns=cols)
df = df.append(results, ignore_index=True)
if df.shape[0]>1:
df = df[df["confidence"] == np.max(df["confidence"])]
return df.iloc[0]["xcenter"], df.iloc[0]["ycenter"], df.iloc[0]["width"], df.iloc[0]["height"], df.iloc[0]["confidence"]
def getBoundsVanilla(img):
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
ret,thresh = cv2.threshold(gray, 5, 255,cv2.THRESH_BINARY)
contours, hierarchy = cv2.findContours(thresh, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
curated_contours = []
for c in contours:
if len(c) > 1000:
curated_contours.append(c)
if not len(curated_contours) >= 1:
curated_contours = contours
stacked = np.vstack(curated_contours)
reshaped_contours = stacked.reshape((stacked.shape[0], 2))
d_x = np.max(reshaped_contours[:, 0]) - np.min(reshaped_contours[:, 0])
d_y = np.max(reshaped_contours[:, 1]) - np.min(reshaped_contours[:, 1])
d = np.max((d_x, d_y))
center = [(np.max(reshaped_contours[:, 0]) + np.min(reshaped_contours[:, 0]))/2, (np.max(reshaped_contours[:, 1]) + np.min(reshaped_contours[:, 1]))/2 ]
x_min = int(center[1] - d/2)
x_max = int(center[1] + d/2)
y_min = int(center[0] - d/2)
y_max = int(center[0] + d/2)
if x_min < y_min and x_min < 0:
x_min = 0
x_max = img.shape[0]
y_min = np.max((0, y_min))
y_max = np.min((y_max, img.shape[1]))
elif y_min < x_min and y_min < 0:
y_min = 0
y_max = img.shape[1]
x_min = np.max((0, x_min))
x_max = np.min((x_max, img.shape[0]))
to_pad = np.abs((x_max+1 - x_min) - (y_max+1 - y_min))
cropped_image = img[x_min:x_max+1, y_min:y_max+1]
if x_max+1 - x_min > y_max+1 - y_min:
padded = cv2.copyMakeBorder(cropped_image, top=0, bottom=0, left=int(to_pad/2), right=int(to_pad/2), borderType=cv2.BORDER_CONSTANT, value=[0,0,0])
else:
padded = cv2.copyMakeBorder(cropped_image, top=int(to_pad/2), bottom=int(to_pad/2), left=0, right=0, borderType=cv2.BORDER_CONSTANT, value=[0,0,0])
return padded
def getODFromCroppedImage(model, image):
#image = cv2.imread(imagePath)
#Crop to 288 for Yolo
im288 = image.resize((288,288))
OD_xcent, OD_ycent, OD_width, OD_height, OD_conf = getOD(model, im288)
#print(OD_xcent, OD_ycent, OD_width, OD_height, OD_conf)
#cropping based on bounds
x_start = OD_xcent - OD_width/2
x_end = x_start + OD_width
y_start = OD_ycent - OD_height/2
y_end = y_start + OD_height
scale = image.size[0]
x_start = int(x_start * scale)
x_end = int(x_end * scale)
y_start = int(y_start * scale)
y_end = int(y_end * scale)
y_diff = y_end - y_start
x_diff = x_end - x_start
if x_diff < y_diff:
x_adj = y_diff - x_diff
left_adj = int(x_adj/2)
right_adj = x_adj - left_adj
x_start -= left_adj
x_end += right_adj
elif y_diff < x_diff:
y_adj = x_diff - y_diff
up_adj = int(y_adj/2)
down_adj = y_adj - up_adj
y_start -= up_adj
y_end += down_adj
OD_fullsize = image.crop((x_start, y_start, x_end, y_end))
return OD_fullsize, OD_conf
def getAutoencoderLoss(image, model, imgSize):
criterion = nn.MSELoss()
transform = transforms.ToTensor()
tensor = transform(image)
resizeTransform = transforms.Resize([imgSize, imgSize])
if GPU:
tensor = resizeTransform(tensor).cuda()
else:
tensor = resizeTransform(tensor).cpu()
if GPU:
model.cuda()
else:
model.cpu()
with torch.no_grad():
model.eval()
output = model(tensor[None, ...])
loss = criterion(output, tensor[None, ...])
return loss.cpu().detach().numpy().flatten()[0]
def getPreds(OD_fullsize, imageCroppedBounds, OD_conf, densenet, seresnext, effnet, vgg16, inception, autoencoder, vae):
densenet_preds = []
vgg16_preds = []
seresnext_preds = []
effnet_preds = []
inception_preds = []
autoencoder_preds = []
vae_preds = []
for flip in ["h", "v", "neither"]:
if flip == "h":
imTransposed = OD_fullsize.transpose(Image.FLIP_TOP_BOTTOM)
imageCroppedBounds_flipped = imageCroppedBounds.transpose(Image.FLIP_TOP_BOTTOM)
elif flip == "v":
imTransposed = OD_fullsize.transpose(Image.FLIP_LEFT_RIGHT)
imageCroppedBounds_flipped = imageCroppedBounds.transpose(Image.FLIP_LEFT_RIGHT)
else:
imTransposed = OD_fullsize
imageCroppedBounds_flipped = imageCroppedBounds
seresnext_pred = predictSingle(seresnext, imTransposed, 224)
densenet_pred = predictSingle(densenet, imTransposed, 120)
vgg16_pred = predictSingle(vgg16, imTransposed, 120)
effnet_pred = predictSingle(effnet, imTransposed, 224)
inception_pred = predictSingle(inception, imTransposed, 299)
seresnext_preds.append(seresnext_pred)
densenet_preds.append(densenet_pred)
vgg16_preds.append(vgg16_pred)
effnet_preds.append(effnet_pred)
inception_preds.append(inception_pred)
autoencoder_loss = getAutoencoderLoss(imageCroppedBounds_flipped, autoencoder, 256)
vae_loss = getAutoencoderLoss(imageCroppedBounds_flipped, vae, 224)
autoencoder_preds.append(autoencoder_loss)
vae_preds.append(vae_loss)
#rg_likelihood = np.mean((seresnext_pred, densenet_pred, vgg16_pred, effnet_pred, inception_pred))
rg_likelihood = np.mean((np.mean(seresnext_preds),
np.mean(densenet_preds),
np.mean(vgg16_preds),
np.mean(effnet_preds),
np.mean(inception_preds)))
rg_binary = bool(rg_likelihood > 0.5)
if rg_likelihood >= 0 and rg_likelihood <= 0.5:
pred_scale = rg_likelihood
else:
pred_scale = -(rg_likelihood - 1)
ungradability_score = pred_scale * np.mean(autoencoder_preds) * np.mean(vae_preds)
ungradability_binary = bool(pred_scale > 0.2)
out = {
"referable-glaucoma-likelihood": rg_likelihood, # Likelihood for 'referable glaucoma'
"referable-glaucoma-binary": rg_binary, # True if 'referable glaucoma', False if 'no referable glaucoma'
"ungradability-score": ungradability_score, # True if 'ungradable', False if 'gradable'
"ungradability-binary": ungradability_binary # The higher the value, the more likely the label is 'ungradable'
}
return out
def getModels(yolo_weights, densenet_weights, seresnext_weights, effnet_weights, vgg_16_weights, inception_weights, autoencoder_weights, vae_weights):
densenet = models.densenet161(pretrained=False)
num_ftrs = densenet.classifier.in_features
densenet.classifier = nn.Sequential(nn.Dropout(0.0), nn.Linear(num_ftrs, 1, bias=True))
if GPU:
densenet.load_state_dict(torch.load(densenet_weights))
else:
densenet.load_state_dict(torch.load(densenet_weights, map_location=torch.device('cpu')))
seresnext = se_resnext50_32x4d(num_classes=1000, pretrained=None)
num_ftrs = seresnext.last_linear.in_features
seresnext.avg_pool = nn.AdaptiveAvgPool2d((1,1))
seresnext.last_linear = nn.Sequential(nn.Dropout(0.0), nn.Linear(num_ftrs, 1, bias=True))
if GPU:
seresnext.load_state_dict(torch.load(seresnext_weights))
else:
seresnext.load_state_dict(torch.load(seresnext_weights, map_location=torch.device('cpu')))
effnet = EfficientNet.from_name('efficientnet-b7')
num_ftrs = effnet._fc.in_features
effnet._fc = nn.Sequential(nn.Dropout(0.0), nn.Linear(num_ftrs, 1, bias=True))
if GPU:
effnet.load_state_dict(torch.load(effnet_weights))
else:
effnet.load_state_dict(torch.load(effnet_weights, map_location=torch.device('cpu')))
vgg16 = models.vgg16()
vgg16.classifier[5] = nn.Dropout(0.0)
num_ftrs = vgg16.classifier[6].in_features
vgg16.classifier[6] = nn.Sequential(nn.Linear(num_ftrs, 1, bias=True))
if GPU:
vgg16.load_state_dict(torch.load(vgg_16_weights))
else:
vgg16.load_state_dict(torch.load(vgg_16_weights, map_location=torch.device('cpu')))
autoencoder = Autoencoder()
if GPU:
autoencoder.load_state_dict(torch.load(autoencoder_weights))
else:
autoencoder.load_state_dict(torch.load(autoencoder_weights, map_location=torch.device('cpu')))
vae = VAE2(224)
if GPU:
vae.load_state_dict(torch.load(vae_weights))
else:
vae.load_state_dict(torch.load(vae_weights, map_location=torch.device('cpu')))
yolo = torch.hub.load(yolo_weights, 'custom', path='yolov5_weights.pt', source="local")
inceptionv3 = models.inception_v3(pretrained=False)
inceptionv3.dropout = nn.Dropout(0.0)
num_ftrs = inceptionv3.fc.in_features
inceptionv3.fc = nn.Linear(num_ftrs, 1, bias=True)
if GPU:
inceptionv3.load_state_dict(torch.load(inception_weights))
else:
inceptionv3.load_state_dict(torch.load(inception_weights, map_location=torch.device('cpu')))
return {"yolo": yolo, "densenet": densenet, "seresnext": seresnext,
"effnet": effnet, "vgg16": vgg16, "inception": inceptionv3,
"autoencoder": autoencoder, "vae": vae}
def predictSingle(model, img, size):
transform = transforms.ToTensor()
tensor = transform(img)
resizeTransform = transforms.Resize([size, size])
tensor = resizeTransform(tensor)
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
if GPU:
tensor = normalize(tensor).cuda()
else:
tensor = normalize(tensor).cpu()
if GPU:
model.cuda()
else:
model.cpu()
with torch.no_grad():
model.eval()
output = torch.sigmoid(model(tensor[None, ...]))
result = output.cpu().detach().numpy().flatten()
return result[0]
def predict(image, models):
image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
imageCroppedBounds = getBoundsVanilla(image)
imageCroppedBounds = cv2.cvtColor(imageCroppedBounds, cv2.COLOR_BGR2RGB)
imageCroppedBounds = Image.fromarray(imageCroppedBounds)
OD_fullsize, OD_conf = getODFromCroppedImage(models["yolo"], imageCroppedBounds)
out = getPreds(OD_fullsize, imageCroppedBounds, OD_conf, models["densenet"], models["seresnext"],
models["effnet"], models["vgg16"], models["inception"], models["autoencoder"], models["vae"])
return out, OD_conf
class DummyLoader(ImageLoader):
@staticmethod
def load_image(fname):
return str(fname)
@staticmethod
def hash_image(image):
return hash(image)
class airogs_algorithm(ClassificationAlgorithm):
def __init__(self):
super().__init__(
validators=dict(
input_image=(
UniqueImagesValidator(),
UniquePathIndicesValidator(),
)
),
)
self._file_loaders = dict(input_image=DummyLoader())
self.output_keys = ["multiple-referable-glaucoma-likelihoods",
"multiple-referable-glaucoma-binary",
"multiple-ungradability-scores",
"multiple-ungradability-binary"]
def load(self):
for key, file_loader in self._file_loaders.items():
fltr = (
self._file_filters[key] if key in self._file_filters else None
)
self._cases[key] = self._load_cases(
folder=Path("/input/images/color-fundus/"),
file_loader=file_loader,
file_filter=fltr,
)
pass
def combine_dicts(self, dicts):
out = {}
for d in dicts:
for k, v in d.items():
if k not in out:
out[k] = []
out[k].append(v)
return out
def process_case(self, *, idx, case):
# Load and test the image(s) for this case
print("----Submission FINAL-----")
print("DEVICE is: " + device)
yolo_weights = 'yolov5'
se_resnext_weights = "se_resnext_weights.pth"
densenet_weights = "densenet_weights.pth"
vgg16_weights = "vgg16_weights.pth"
effnet_weights = "effnetb7_weights.pth"
inception_weights = "inceptionv3_weights.pth"
autoencoder_weights = "autoencoder_weights.pth"
vae_weights = "vae_weights.pth"
my_models = getModels(yolo_weights, densenet_weights, se_resnext_weights, effnet_weights, vgg16_weights, inception_weights, autoencoder_weights, vae_weights)
if case.path.suffix == '.tiff':
results = []
with tifffile.TiffFile(case.path) as stack:
for page in tqdm.tqdm(stack.pages):
input_image_array = page.asarray()
results.append(self.predict(my_models=my_models, input_image_array=input_image_array))
else:
input_image = SimpleITK.ReadImage(str(case.path))
input_image_array = SimpleITK.GetArrayFromImage(input_image)
results = [self.predict(input_image_array=input_image_array)]
results = self.combine_dicts(results)
# Test classification output
if not isinstance(results, dict):
raise ValueError("Expected a dictionary as output")
return results
def predict(self, *, my_models: Dict, input_image_array: np.ndarray) -> Dict:
results, OD_conf = predict(input_image_array, my_models)
rg_likelihood = float(results["referable-glaucoma-likelihood"])
rg_binary = results["referable-glaucoma-binary"]
ungradability_score = float(results["ungradability-score"])
ungradability_binary = results["ungradability-binary"]
out = {
"multiple-referable-glaucoma-likelihoods": rg_likelihood,
"multiple-referable-glaucoma-binary": rg_binary,
"multiple-ungradability-scores": ungradability_score,
"multiple-ungradability-binary": ungradability_binary
}
return out
def save(self):
for key in self.output_keys:
with open(f"/output/{key}.json", "w") as f:
out = []
for case_result in self._case_results:
out += case_result[key]
json.dump(out, f)
if __name__ == "__main__":
airogs_algorithm().process()