-
Notifications
You must be signed in to change notification settings - Fork 0
/
method1.py
288 lines (211 loc) · 8.5 KB
/
method1.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
import sys
import os
import torch
from PIL import Image, ImageFilter
import numpy as np
from sklearn.mixture import GaussianMixture as GMM
# local
from model import TextExtractor
from lib import extract_patches, merge_patches
def extract_text(in_im, model, patch_size, stride, device):
'''
run the model with all the patches
'''
in_im = in_im.transpose((2, 0, 1))
# the models expects patches so extract them and
# that too in channel first order.
patches = extract_patches(in_im, patch_size, stride)
patches = torch.from_numpy(patches).to(device)
out_patches = model(patches)
out_im = merge_patches(out_patches.detach().cpu().numpy(), in_im.shape[1:],
stride)
return out_im
def extract_text_serial(in_im, model, patch_size, stride, device):
'''
restore patch by patch
reduces the memory requirement
'''
im = in_im.transpose((2, 0, 1))
patch_r, patch_c = patch_size
stride_r, stride_c = stride
nch, nrow, ncol = im.shape
row_idxs = range(0, nrow-patch_r+1, stride_r)
col_idxs = range(0, ncol-patch_c+1, stride_c)
out_im = np.zeros((nrow, ncol, nch), dtype='float32')
count = np.zeros((nrow, ncol), dtype='int')
numpatch = len(row_idxs)*len(col_idxs)
patch_idx = 0
for i in row_idxs:
for j in col_idxs:
patch = im[:, i:i+patch_r, j:j+patch_c]
patch = torch.from_numpy(patch[None, ...]).to(device)
# I am not sure why this float() is necessary
# check
out_patch = model(patch.float())
# out_patch ~> (1, nch, patch_r, patch_c)
out_patch = out_patch.detach().cpu().numpy()
out_im[i:i+patch_r, j:j+patch_c, :] += \
out_patch[0].transpose((1, 2, 0))
count[i:i+patch_r, j:j+patch_c] += 1
sys.stdout.write('Progress: [{}/{}]\r'.format(patch_idx,
numpatch))
patch_idx += 1
sys.stdout.write('\n')
nz = (count != 0)
nz_r = np.tile(nz[..., None], (1, 1, nch))
count_r = np.tile(count[..., None], (1, 1, nch))
out_im[nz_r] = out_im[nz_r] / count_r[nz_r]
return out_im
def extract_text_mini_batch(in_im, model, patch_size, stride,
batch_size, device):
'''
A balance between the above two
'''
im = in_im.transpose((2, 0, 1))
patch_r, patch_c = patch_size
stride_r, stride_c = stride
nch, nrow, ncol = im.shape
row_idxs = range(0, nrow-patch_r+1, stride_r)
col_idxs = range(0, ncol-patch_c+1, stride_c)
col_v, row_v = np.meshgrid(col_idxs, row_idxs)
col_v = col_v.flatten()
row_v = row_v.flatten()
out_im = np.zeros((nrow, ncol, nch), dtype='float32')
count = np.zeros((nrow, ncol), dtype='int')
patch_batch = np.zeros((batch_size, nch) + patch_size, dtype='float32')
numpatch = len(row_idxs)*len(col_idxs)
num_batch = numpatch // batch_size
# handle the non full sized batch separately
for b in range(num_batch):
b_i = b*batch_size
b_j = b*batch_size
for b_idx in range(batch_size):
i = row_v[b_i + b_idx]
j = col_v[b_j + b_idx]
patch_batch[b_idx] = im[:, i:i+patch_r, j:j+patch_c]
patch_batch_t = torch.from_numpy(patch_batch).to(device)
with torch.no_grad():
out_patch = model(patch_batch_t)
# out_patch ~> (num_batch, nch, patch_r, patch_c)
out_patch_b = out_patch.detach().cpu().numpy()
for b_idx in range(batch_size):
i = row_v[b_i + b_idx]
j = col_v[b_j + b_idx]
out_im[i:i+patch_r, j:j+patch_c, :] += \
out_patch_b[b_idx].transpose((1, 2, 0))
count[i:i+patch_r, j:j+patch_c] += 1
sys.stdout.write('Progress: [{}/{}]\r'.format(b, num_batch))
sys.stdout.write('\n')
# handle the non full batch
if numpatch % batch_size != 0:
remaining_patches = numpatch % batch_size
patch_batch = np.zeros((remaining_patches, 1) + patch_size,
dtype='float32')
b_i = num_batch*batch_size
b_j = num_batch*batch_size
for b_idx in range(remaining_patches):
i = row_v[b_i + b_idx]
j = col_v[b_j + b_idx]
patch_batch[b_idx] = im[:, i:i+patch_r, j:j+patch_c]
patch_batch_t = torch.from_numpy(patch_batch).to(device)
with torch.no_grad():
out_patch = model(patch_batch_t)
# out_patch ~> (num_batch, nch, patch_r, patch_c)
out_patch_b = out_patch.detach().cpu().numpy()
for b_idx in range(remaining_patches):
i = row_v[b_i + b_idx]
j = col_v[b_j + b_idx]
out_im[i:i+patch_r, j:j+patch_c, :] += \
out_patch_b[b_idx].transpose((1, 2, 0))
count[i:i+patch_r, j:j+patch_c] += 1
nz = (count != 0)
nz_r = np.tile(nz[..., None], (1, 1, nch))
count_r = np.tile(count[..., None], (1, 1, nch))
out_im[nz_r] = out_im[nz_r] / count_r[nz_r]
return out_im
def restore_background(in_im, in_im_PIL, GMM_k=4):
X = np.reshape(in_im, (-1, in_im.shape[-1]))
clf = GMM(n_components=GMM_k, covariance_type='full')
clf.fit(X)
mu = clf.means_
covar = clf.covariances_
# # convert to gray the means
if in_im_PIL.mode == 'L':
gray_mu = mu
else:
gray_mu = 0.3*mu[:, 0] + 0.59*mu[:, 1] + 0.1*mu[:, 2]
# highest intensity that is not close to 1 is the bg color
if np.sum(in_im > 0.9):
gray_mu[gray_mu.argmax()] = 0
bg_idx = np.argmax(gray_mu)
# generate background
bg = np.random.multivariate_normal(mu[bg_idx], covar[bg_idx],
in_im.shape[:-1])
bg_PIL = Image.fromarray((bg*255).astype('uint8'))
bg_smooth = np.array(bg_PIL.filter(
ImageFilter.GaussianBlur(radius=5)))
return bg_smooth
def restore_image(in_im_PIL, text_extractor_net, patch_size, stride, device,
inference_batch_size=None):
if in_im_PIL.mode != 'RGB' or in_im_PIL.mode != 'L':
in_im_PIL = in_im_PIL.convert('RGB')
in_im_uint8 = np.asarray(in_im_PIL)
in_im = in_im_uint8/255.0
if in_im_PIL.mode == 'L':
in_im_gray = in_im
else:
in_im_gray = np.array(in_im_PIL.convert('L'))/255.0
in_im_gray = np.expand_dims(in_im_gray, -1)
# The trained CNN requires a grayscale image
# main method
# # extract text
if inference_batch_size is None:
out_im = extract_text(in_im_gray, text_extractor_net, patch_size,
stride, device)
else:
text_im = extract_text_mini_batch(in_im_gray, text_extractor_net,
patch_size, stride,
inference_batch_size, device)
# Threshold selection
text_im_uint8 = (text_im*255).astype('uint8')
# bincount requires 1D array
hist = np.bincount(text_im_uint8.flatten(), minlength=256)
smoothed_hist = np.convolve(hist, np.ones((5, ))/5, mode='valid')
thr = np.argmin(smoothed_hist)+2
text_mask = text_im_uint8 < thr
# get background color
bg_smooth = restore_background(in_im, in_im_PIL)
# overlay
# foreground color restoration
fg = in_im_uint8*text_mask
# both fg and bg_smooth are uint8
out_im = fg + (1-text_mask)*bg_smooth
out_im_PIL = Image.fromarray(np.squeeze(out_im).astype('uint8'))
return out_im_PIL
if __name__ == '__main__':
model_wt = './model/upto2017_model_ourdata.pt'
patch_size = (256, 256)
stride = (10, 10)
inference_batch_size = 80
if len(sys.argv) < 2 or len(sys.argv) > 3:
print("Usage: python <code.py> input_image [output_location]")
print("Output is saved in the currnet path if path"
+ " is not provided")
sys.exit(0)
input_im_path = sys.argv[1]
input_file_name, ext = os.path.splitext(os.path.basename(input_im_path))
if len(sys.argv) == 3:
output_im_loc = sys.argv[2]
else:
output_im_loc = '.'
output_im_path = os.path.join(output_im_loc, input_file_name+"_out"+ext)
in_im_PIL = Image.open(input_im_path)
# inference on cpu
device = torch.device('cpu')
model = TextExtractor()
model.load_state_dict(torch.load(model_wt, map_location=device))
model.eval()
out_im_PIL = restore_image(in_im_PIL, model, patch_size, stride, device,
inference_batch_size)
# save output
out_im_PIL.save(output_im_path)