-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathssdd_val.py
155 lines (139 loc) · 5.79 KB
/
ssdd_val.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
import datetime
import math
import os
import random
import re
import cv2
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torch.utils.data
from torch.autograd import Variable
import imutils
import utils
from torchvision import transforms
from torch.utils.data.dataloader import default_collate
import time
from PIL import Image
from base_class import BaseModel, SegBaseModel, SSDDBaseModel, PascalDataset
from network import SegmentationPsa, PredictDiff, PredictDiffHead
import math
############################################################
# dataset
############################################################
class SSDDValData(PascalDataset):
def __init__(self, dataset, config):
super().__init__(dataset, config)
self.joint_transform_list=[
imutils.Rescale(self.config.INP_SHAPE),
None,
None,
]
self.img_transform_list=[
np.asarray,
imutils.Normalize(mean = self.mean, std = self.std),
imutils.HWC_to_CHW
]
def __getitem__(self, image_index):
image_id = self.image_ids[image_index]
# Load image and mask
impath= self.config.VOC_ROOT+'/JPEGImages/'
imn=impath+image_id+'.jpg'
img = Image.open(imn).convert("RGB")
img = self.img_label_resize([img])[0]
images = torch.from_numpy(img)
return images, image_index
def __len__(self):
return self.image_ids.shape[0]
############################################################
# Model Class
############################################################
class SegModel(SegBaseModel):
def __init__(self, config):
super(SegModel, self).__init__(config)
in_channel=4096
self.seg_main = SegmentationPsa(config, in_channel=in_channel, middle_channel=512, num_classes=21)
def forward(self, inputs):
x = inputs
[x1, x2, x3, x4, x5] = self.encoder(x)
seg, seg_head = self.seg_main(x5)
return seg
class Evaluator():
def __init__(self, config, model):
super(Evaluator, self).__init__()
self.config = config
self.model=model
def eval_model(self, val_dataset):
self.val_set = SSDDValData(val_dataset, self.config)
val_generator = torch.utils.data.DataLoader(self.val_set, batch_size=self.config.BATCH, shuffle=False, num_workers=torch.cuda.device_count()*2, pin_memory=True)
self.model.eval()
self.eval(val_generator)
def get_segmentation(self, img):
segs = self.get_ms_segout(img)
fimg = img[:,:,:,torch.arange(img.shape[3]-1,-1,-1)]
fsegs = self.get_ms_segout(fimg)
seg_all = torch.zeros(1,segs[0].shape[1],segs[0].shape[2],segs[0].shape[3])
for i in range(len(segs)):
seg_all += segs[i]
for i in range(len(segs)):
seg_all += fsegs[i][:,:,:,torch.arange(fsegs[i].shape[3]-1,-1,-1)]
return seg_all
def get_ms_segout(self, img):
scales = [1/2, 3/4, 1, 5/4, 3/2]
segs = []
for i in range(len(scales)):
scale=scales[i]
simg = F.interpolate(img, (int(img.shape[2]*scale),int(img.shape[3]*scale)), mode='bilinear')
seg = self.model(simg)
seg = F.softmax(seg,dim=1)
seg = F.interpolate(seg, (int(img.shape[2]),int(img.shape[3])), mode='bilinear')
seg = seg.data.cpu()
segs.append(seg)
torch.cuda.empty_cache()
return segs
def eval(self, datagenerator):
end = time.time()
cnt=0
for inputs in datagenerator:
print(cnt)
data_time = time.time()
start=time.time()
images, imgindex = inputs
images = Variable(images).cuda()
segs=[]
with torch.no_grad():
for i in range(len(images)):
# segmentation
seg=self.get_segmentation(images[i:i+1])
# crf
image_id = self.val_set.image_ids[imgindex[i]]
impath=self.config.VOC_ROOT+'/JPEGImages/'
imn=impath+image_id+'.jpg'
img_org = np.asarray(Image.open(imn))
seg=F.interpolate(seg,(img_org.shape[0],img_org.shape[1]),mode='bilinear')
prob=F.softmax(seg,dim=1)[0].data.cpu().numpy()
seg_mask = np.argmax(prob,0)
seg_crf_map = imutils.crf_inference(img_org, prob, labels=prob.shape[0], t=10)
seg_crf_mask = np.argmax(seg_crf_map,axis=0)
# save results
cnt+=1
saven = os.path.join(self.savedir, 'seg_val_'+self.saveid+'_'+str(cnt)+'.png')
utils.mask2png(saven, seg_mask)
saven = os.path.join(self.savedir, 'seg_val_'+self.saveid+'_'+str(cnt)+'.txt')
np.savetxt(saven, seg_mask)
saven = os.path.join(self.savedir, 'seg_val_crf_'+self.saveid+'_'+str(cnt)+'.png')
utils.mask2png(saven, seg_crf_mask)
saven = os.path.join(self.savedir, 'seg_val_crf_'+self.saveid+'_'+str(cnt)+'.txt')
np.savetxt(saven, seg_crf_mask)
def set_log_dir(self, phase, saveid, model_path=None):
self.phase = phase
self.saveid = saveid
self.savedir = 'validation/'+self.saveid
print("save the results to "+self.savedir)
if not os.path.exists(self.savedir):
os.makedirs(self.savedir)
def val(config, weight_file=None):
model = SegModel(config=config)
return model