forked from axinc-ai/ailia-models
-
Notifications
You must be signed in to change notification settings - Fork 0
/
illnet.py
188 lines (149 loc) · 5.35 KB
/
illnet.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
import sys
import time
import numpy as np
import cv2
from skimage import img_as_ubyte
from skimage import io
import ailia
from illnet_utils import *
# import original modules
sys.path.append('../../util')
from utils import get_base_parser, update_parser # noqa: E402
from model_utils import check_and_download_models # noqa: E402
import webcamera_utils # noqa: E402
# ======================
# Parameters
# ======================
WEIGHT_PATH = 'illnet.onnx'
MODEL_PATH = 'illnet.onnx.prototxt'
REMOTE_PATH = 'https://storage.googleapis.com/ailia-models/illnet/'
IMAGE_PATH = 'input.png'
SAVE_IMAGE_PATH = 'output.png'
PATCH_RES = 128
# ======================
# Arguemnt Parser Config
# ======================
parser = get_base_parser(
'Illumination Correction Model', IMAGE_PATH, SAVE_IMAGE_PATH
)
args = update_parser(parser)
# ======================
# Main functions
# ======================
def recognize_from_image():
# prepare input data
img = io.imread(args.input)
img = preProcess(img)
input_data = padCropImg(img)
input_data = input_data.astype(np.float32) / 255.0
ynum = input_data.shape[0]
xnum = input_data.shape[1]
preds_ailia = np.zeros(
(ynum, xnum, PATCH_RES, PATCH_RES, 3), dtype=np.float32
)
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
# inference
print('Start inference...')
if args.benchmark:
print('BENCHMARK mode')
for c in range(5):
start = int(round(time.time() * 1000))
for j in range(ynum):
for i in range(xnum):
patchImg = input_data[j, i]
patchImg = (patchImg - 0.5) / 0.5
patchImg = patchImg.transpose((2, 0, 1))
patchImg = patchImg[np.newaxis, :, :, :]
out = net.predict(patchImg)
out = out.transpose((0, 2, 3, 1))[0]
out = (np.clip(out, 0, 1) * 255).astype(np.uint8)
preds_ailia[j, i] = out
end = int(round(time.time() * 1000))
print(f'\tailia processing time {end - start} ms')
else:
start = int(round(time.time() * 1000))
for j in range(ynum):
for i in range(xnum):
patchImg = input_data[j, i]
patchImg = (patchImg - 0.5) / 0.5
patchImg = patchImg.transpose((2, 0, 1))
patchImg = patchImg[np.newaxis, :, :, :]
out = net.predict(patchImg)
out = out.transpose((0, 2, 3, 1))[0]
out = (np.clip(out, 0, 1) * 255).astype(np.uint8)
preds_ailia[j, i] = out
end = int(round(time.time() * 1000))
# postprocessing
resImg = composePatch(preds_ailia)
resImg = postProcess(resImg)
resImg.save(args.savepath)
print('Script finished successfully.')
def recognize_from_video():
# [WARNING] This is test impl
print('[WARNING] This is test implementation')
# net initialize
net = ailia.Net(MODEL_PATH, WEIGHT_PATH, env_id=args.env_id)
capture = webcamera_utils.get_capture(args.video)
# create video writer if savepath is specified as video format
if args.savepath != SAVE_IMAGE_PATH:
f_h = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))
f_w = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH))
dummy_img = np.zeros((f_h, f_w, 3))
dummy_img = padCropImg(dummy_img)
ynum = dummy_img.shape[0]
xnum = dummy_img.shape[1]
dummy_img = np.zeros(
(ynum, xnum, PATCH_RES, PATCH_RES, 3), dtype=np.float32
)
dummy_img = composePatch(dummy_img)
writer = webcamera_utils.get_writer(
args.savepath, dummy_img.shape[0], dummy_img.shape[1]
)
else:
writer = None
while(True):
ret, frame = capture.read()
if (cv2.waitKey(1) & 0xFF == ord('q')) or not ret:
break
img = preProcess(frame)
input_data = padCropImg(img)
input_data = input_data.astype(np.float32) / 255.0
ynum = input_data.shape[0]
xnum = input_data.shape[1]
preds_ailia = np.zeros(
(ynum, xnum, PATCH_RES, PATCH_RES, 3), dtype=np.float32
)
for j in range(ynum):
for i in range(xnum):
patchImg = input_data[j, i]
patchImg = (patchImg - 0.5) / 0.5
patchImg = patchImg.transpose((2, 0, 1))
patchImg = patchImg[np.newaxis, :, :, :]
out = net.predict(patchImg)
out = out.transpose((0, 2, 3, 1))[0]
out = (np.clip(out, 0, 1) * 255).astype(np.uint8)
preds_ailia[j, i] = out
resImg = composePatch(preds_ailia)
resImg = postProcess(resImg)
resImg = img_as_ubyte(resImg)
cv2.imshow('frame', resImg)
# save results
if writer is not None:
writer.write(resImg)
capture.release()
cv2.destroyAllWindows()
if writer is not None:
writer.release()
print('Script finished successfully.')
def main():
# model files check and download
check_and_download_models(WEIGHT_PATH, MODEL_PATH, REMOTE_PATH)
if args.video is not None:
# video mode
recognize_from_video()
else:
# image mode
recognize_from_image()
if __name__ == '__main__':
main()