-
Notifications
You must be signed in to change notification settings - Fork 9
/
demo.py
executable file
·137 lines (110 loc) · 6.6 KB
/
demo.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
import string
import argparse
import torch
import torch.backends.cudnn as cudnn
import torch.utils.data
import torch.nn.functional as F
import numpy as np
from utils import CTCLabelConverter, AttnLabelConverter, TransLabelConverter
from dataset import RawDataset, AlignCollate
from model import Model
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
def demo(opt):
""" model configuration """
converter = AttnLabelConverter(opt.character)
opt.num_class = len(converter.character)
if opt.rgb:
opt.input_channel = 3
model = Model(opt)
# print('model input parameters', opt.imgH, opt.imgW, opt.num_fiducial, opt.input_channel, opt.output_channel,
# opt.hidden_size, opt.num_class, opt.batch_max_length, opt.Transformation, opt.FeatureExtraction,
# opt.SequenceModeling, opt.Prediction)
model = torch.nn.DataParallel(model).to(device)
# load model
print('loading pretrained model from %s' % opt.saved_model)
model.load_state_dict(torch.load(opt.saved_model, map_location=device))
# prepare data. two demo images from https://github.com/bgshih/crnn#run-demo
AlignCollate_demo = AlignCollate(imgH=opt.imgH, imgW=opt.imgW, keep_ratio_with_pad=opt.PAD)
demo_data = RawDataset(root=opt.image_folder, opt=opt) # use RawDataset
demo_loader = torch.utils.data.DataLoader(
demo_data, batch_size=opt.batch_size,
shuffle=False,
num_workers=int(opt.workers),
collate_fn=AlignCollate_demo, pin_memory=True)
# predict
model.eval()
with torch.no_grad():
for image_tensors, image_path_list in demo_loader:
all_pred_strs = []
all_confidence_scores = []
batch_size = image_tensors.size(0)
image = image_tensors.to(device)
# For max length prediction
length_for_pred = torch.IntTensor([opt.batch_max_length] * batch_size).to(device)
text_for_pred = torch.LongTensor(batch_size, opt.batch_max_length + 1).fill_(0).to(device)
predss = model(image, text_for_pred, is_train=False)[0]
for i, preds in enumerate(predss):
confidence_score_list = []
pred_str_list = []
# select max probability (greedy decoding) then decode index to character
_, preds_index = preds.max(2)
preds_str = converter.decode(preds_index, length_for_pred)
preds_prob = F.softmax(preds, dim=2)
preds_max_prob, _ = preds_prob.max(dim=2)
for pred, pred_max_prob in zip(preds_str, preds_max_prob):
pred_EOS = pred.find('[s]')
pred = pred[:pred_EOS] # prune after "end of sentence" token ([s])
pred_str_list.append(pred)
pred_max_prob = pred_max_prob[:pred_EOS]
# calculate confidence score (= multiply of pred_max_prob)
try:
confidence_score = pred_max_prob.cumprod(dim=0)[-1].cpu().numpy()
except:
confidence_score = 0 # for empty pred case, when prune after "end of sentence" token ([s])
confidence_score_list.append(confidence_score)
all_pred_strs.append(pred_str_list)
all_confidence_scores.append(confidence_score_list)
all_confidence_scores = np.array(all_confidence_scores)
all_pred_strs = np.array(all_pred_strs)
best_pred_index = np.argmax(all_confidence_scores, axis=0)
best_pred_index = np.expand_dims(best_pred_index, axis=0)
# Get max predition per image through blocks
all_pred_strs = np.take_along_axis(all_pred_strs, best_pred_index, axis=0)[0]
all_confidence_scores = np.take_along_axis(all_confidence_scores, best_pred_index, axis=0)[0]
log = open(f'./log_demo_result.txt', 'w')
dashed_line = '-' * 80
head = f'{"image_path":25s}\t{"predicted_labels":25s}\tconfidence score'
print(f'{dashed_line}\n{head}\n{dashed_line}')
log.write(f'{dashed_line}\n{head}\n{dashed_line}\n')
for img_name, pred, confidence_score in zip(image_path_list, all_pred_strs, all_confidence_scores):
print(f'{img_name:25s}\t{pred:25s}\t{confidence_score:0.4f}')
log.write(f'{img_name:25s}\t{pred:25s}\t{confidence_score:0.4f}\n')
log.close()
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--image_folder', default='test_image/normal_cases', required=False, help='path to image_folder which contains text images')
parser.add_argument('--workers', type=int, help='number of data loading workers', default=4)
parser.add_argument('--batch_size', type=int, default=4, help='input batch size')
parser.add_argument('--saved_model', default='weights/scatter-case-sensitive.pth', required=False, help="path to saved_model to evaluation")
""" Data processing """
parser.add_argument('--batch_max_length', type=int, default=25, help='maximum-label-length')
parser.add_argument('--imgH', type=int, default=32, help='the height of the input image')
parser.add_argument('--imgW', type=int, default=100, help='the width of the input image')
parser.add_argument('--rgb', action='store_true', help='use rgb input')
parser.add_argument('--character', type=str, default = '''0123456789abcdefghijklmnopqrstuvwxyz''', help='character label')
parser.add_argument('--sensitive', action='store_true', help='for sensitive character mode')
parser.add_argument('--PAD', action='store_true', help='whether to keep ratio then pad for image resize')
""" Model Architecture """
parser.add_argument('--num_fiducial', type=int, default=20, help='number of fiducial points of TPS-STN')
parser.add_argument('--input_channel', type=int, default=1, help='the number of input channel of Feature extractor')
parser.add_argument('--output_channel', type=int, default=512,
help='the number of output channel of Feature extractor')
parser.add_argument('--hidden_size', type=int, default=512, help='the size of the LSTM hidden state')
opt = parser.parse_args()
""" vocab / character number configuration """
if opt.sensitive:
opt.character = string.printable[:-6] # same with ASTER setting (use 94 char).
cudnn.benchmark = True
cudnn.deterministic = True
opt.num_gpu = torch.cuda.device_count()
demo(opt)