-
Notifications
You must be signed in to change notification settings - Fork 0
/
visualization_wsi.py
139 lines (119 loc) · 6.07 KB
/
visualization_wsi.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
import torch
import glob
from model.network import Classifier_1fc, DimReduction
from model.Attention import Attention_Gated as Attention
# from model.Attention import Attention_with_Classifier
import argparse
# from dataset.EmbededFeatsDataset import EmbededFeatsDataset
# torch.autograd.set_detect_anomaly(True)
from sklearn.metrics import roc_auc_score,f1_score,roc_curve
import numpy as np
from utils import eval_metric
from dataset.psemix_core import augment_bag
from PIL import Image
import cv2
import openslide
import cmap
import matplotlib.pyplot as plt
from matplotlib import cm
colormap = cm.get_cmap('jet')
# print(colormap(0.23))
# plt.scatter([x/256 for x in range(256)],[x/256 for x in range(256)],c=[x/256 for x in range(256)],cmap=colormap)
# plt.colorbar()
# plt.savefig('colorbar.png',dpi=600)
parser = argparse.ArgumentParser(description='abc')
parser.add_argument('--name', default='abc', type=str)
parser.add_argument('--EPOCH', default=200, type=int)
parser.add_argument('--epoch_step', default='[100]', type=str)
parser.add_argument('--device', default='cuda', type=str)
parser.add_argument('--isPar', default=False, type=bool)
parser.add_argument('--log_dir', default='./debug_log', type=str) ## log file path
parser.add_argument('--train_show_freq', default=40, type=int)
parser.add_argument('--droprate', default='0', type=float)
parser.add_argument('--droprate_2', default='0', type=float)
parser.add_argument('--lr', default=1e-5, type=float)
parser.add_argument('--weight_decay', default=1e-4, type=float)
parser.add_argument('--lr_decay_ratio', default=0.2, type=float)
parser.add_argument('--batch_size', default=1, type=int)
parser.add_argument('--batch_size_v', default=1, type=int)
parser.add_argument('--num_workers', default=4, type=int)
parser.add_argument('--num_cls', default=2, type=int)
parser.add_argument('--mDATA0_dir_train0', default='', type=str) ## Train Set
parser.add_argument('--mDATA0_dir_val0', default='', type=str) ## Validation Set
parser.add_argument('--mDATA_dir_test0', default='', type=str) ## Test Set
parser.add_argument('--numGroup', default=5, type=int)
parser.add_argument('--total_instance', default=4, type=int)
parser.add_argument('--numGroup_test', default=4, type=int)
parser.add_argument('--total_instance_test', default=4, type=int)
parser.add_argument('--mDim', default=512, type=int)
parser.add_argument('--grad_clipping', default=5, type=float)
parser.add_argument('--isSaveModel', action='store_false')
parser.add_argument('--debug_DATA_dir', default='', type=str)
parser.add_argument('--numLayer_Res', default=0, type=int)
parser.add_argument('--temperature', default=1, type=float)
parser.add_argument('--num_MeanInference', default=1, type=int)
parser.add_argument('--distill_type', default='AFS', type=str) ## MaxMinS, MaxS, AFS
params = parser.parse_args()
wsi_path='/path/to/CAMELYON16/extracted_patches_0.8/testing/images/256.1/test_040'
mask='/path/to/CAMELYON16/mask/test_040.tif'
mask_img=openslide.open_slide(mask)
mask_img=np.array(mask_img.read_region((0,0),4,mask_img.level_dimensions[4]).convert('RGB'))
patches_embedded=np.load(wsi_path+'/axgated_em0_resnet1024_feats.npy')
# patches_embedded=np.load(wsi_path+'/resnet1024_feats.npy')
# patches_embedded=np.load(wsi_path+'/em0_resnet1024_feats.npy')
patches_path=sorted(glob.glob(wsi_path+'/*.png'))
device=torch.device("cuda" if torch.cuda.is_available() else "cpu")
classifier = Classifier_1fc(params.mDim, params.num_cls, params.droprate).to(params.device)
attention = Attention(params.mDim).to(params.device)
dimReduction = DimReduction(1024, params.mDim, numLayer_Res=params.numLayer_Res).to(params.device)
# attCls = Attention_with_Classifier(L=params.mDim, num_cls=params.num_cls, droprate=params.droprate_2).to(params.device)
pretrained_weights=torch.load('AB-MIL_psemix_model_best.pth')
# pretrained_weights=torch.load('AB-MIL_model_best.pth')
# pretrained_weights=torch.load('AB-MIL_EM_model_best.pth')
# pretrained_weights=torch.load('/home/why/Workspace-Python/EM-MIL/AB-MIL_model_best_em0_resnet1024.pth')
classifier.load_state_dict(pretrained_weights['classifier'])
dimReduction.load_state_dict(pretrained_weights['dim_reduction'])
attention.load_state_dict(pretrained_weights['attention'])
classifier.eval()
dimReduction.eval()
attention.eval()
def min_max_norm(tAA):
return (tAA-torch.min(tAA))/(torch.max(tAA)-torch.min(tAA))
patches_embedded=torch.from_numpy(patches_embedded).to(params.device).unsqueeze(0)
with torch.no_grad():
tmidFeat = dimReduction(patches_embedded).squeeze(0)
tAA = attention(tmidFeat,isNorm=False).squeeze(0)
tAA=min_max_norm(tAA)
last_patch=patches_path[-1]
max_col=int(last_patch.split(r'_')[-6])
max_row=int(last_patch.split(r'_')[-5])
shape_col=int(last_patch.split(r'_')[-3])
shape_row=int(last_patch.split(r'_')[-1].split(r'.')[0])
final_output=np.ones((shape_row*20,shape_col*20,3))*255
mask_img=cv2.resize(mask_img,(shape_col*20-60,shape_row*20-120))
mask_img=cv2.copyMakeBorder(mask_img, 60, 60, 30, 30, cv2.BORDER_CONSTANT, None, 0)
edges=cv2.Canny(mask_img, 127, 200)
kernel = np.ones((3,3), np.uint8)
edges =cv2.dilate(edges, kernel)
edges_red=np.zeros_like(final_output)
edges_red[edges!=0]=[0,0,255]
edges_white=np.zeros_like(final_output)
edges_white[edges!=0]=[255,255,255]
idx=0
for patch_path in patches_path:
col=int(patch_path.split(r'_')[-6])+1
row=int(patch_path.split(r'_')[-5])+7
attention_score=tAA[idx].cpu().item()
print(attention_score)
# attention_score=0 if attention_score<5e-17 else 1000*attention_score
idx+=1
patch_img=cv2.imread(patch_path)
patch_img=cv2.resize(patch_img,(20,20))
mask_img= np.array(Image.new("RGB",(20,20),(int(255*colormap(attention_score)[2]),int(255*colormap(attention_score)[1]),int(255*colormap(attention_score)[0]))))
# print(colormap(attention_score))
output_img=0.6*mask_img+0.4*patch_img
final_output[20*row-6:20*row+20-6, 20*col-7:20*col+20-7, :]=output_img
final_output[edges_white==255]=0
cv2.imwrite('040_gated_cmap_fullred.png',final_output+edges_red)
# patch=PIL.Image.open(patch_path)
# patch=torch.from_numpy(np.array(patch)).permute(2, 0, 1).float()/255.0