VITON/model_zalando_refine_test.py at master · xthan/VITON

History

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

#

# Licensed under the Apache License, Version 2.0 (the "License");

# you may not use this file except in compliance with the License.

# You may obtain a copy of the License at

#

# http://www.apache.org/licenses/LICENSE-2.0

#

# Unless required by applicable law or agreed to in writing, software

# distributed under the License is distributed on an "AS IS" BASIS,

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

# See the License for the specific language governing permissions and

# limitations under the License.

# ==============================================================================

""" Test for Stage 2: from product image + warpped image => refined image.

"""

from __future__ import absolute_import

from __future__ import division

from __future__ import print_function

from utils import *

import collections

from model_zalando_tps_warp import create_refine_generator

import os

import time

from tps_transformer import tps_stn

import numpy as np

import scipy.io as sio

import scipy.misc

import tensorflow as tf

from PIL import Image

FLAGS = tf.app.flags.FLAGS

tf.flags.DEFINE_string("image_dir", "data/women_top/",

"Directory containing product and person images.")

tf.flags.DEFINE_string("test_label",

"data/viton_test_pairs.txt",

"File containing labels for testing.")

tf.flags.DEFINE_string("result_dir", "results/stage2/",

"Folder containing the results of testing.")

tf.flags.DEFINE_string("coarse_result_dir", "results/stage1",

"Folder containing the results of stage1 (coarse) results.")

tf.flags.DEFINE_integer("begin", "0", "")

tf.flags.DEFINE_integer("end", "2032", "")

tf.logging.set_verbosity(tf.logging.INFO)

def deprocess_image(image, mask01=False):

if not mask01:

image = image / 2 + 0.5

return image

def process_one_image(image, resize_height, resize_width, if_zero_one=False):

image = tf.image.convert_image_dtype(image, dtype=tf.float32)

if if_zero_one:

return image

image = tf.image.resize_images(image,

size=[resize_height, resize_width],

method=tf.image.ResizeMethod.BILINEAR)

return (image - 0.5) * 2.0

# preprocess images for testing

def _process_image(image_name, product_image_name, sess,

resize_width=192, resize_height=256):

image_id = image_name[:-4]

image = scipy.misc.imread(FLAGS.image_dir + image_name)

prod_image = scipy.misc.imread(FLAGS.image_dir + product_image_name)

# sorry for the hard coded file path.

coarse_image = scipy.misc.imread(FLAGS.coarse_result_dir +

"/images/00015000_" +

image_name + "_" +

product_image_name + ".png")

mask_output = scipy.misc.imread(FLAGS.coarse_result_dir +

"/images/00015000_" +

image_name + "_" +

product_image_name + "_mask.png")

image = process_one_image(image, resize_height, resize_width)

prod_image = process_one_image(prod_image, resize_height, resize_width)

coarse_image = process_one_image(coarse_image, resize_height, resize_width)

mask_output = process_one_image(mask_output, resize_height,

resize_width, True)

# TPS transform

# Here we use control points to generate

# We tried to learn the control points, but the network refuses to converge.

tps_control_points = sio.loadmat(FLAGS.coarse_result_dir +

"/tps/00015000_" +

image_name + "_" +

product_image_name +

"_tps.mat")

v = tps_control_points["control_points"]

nx = v.shape[1]

ny = v.shape[2]

v = np.reshape(v, -1)

v = np.transpose(v.reshape([1,2,nx*ny]), [0,2,1]) * 2 -1

p = tf.convert_to_tensor(v, dtype=tf.float32)

img = tf.reshape(prod_image, [1,256,192,3])

tps_image = tps_stn(img, nx, ny, p, [256,192,3])

tps_mask = tf.cast(tf.less(tf.reduce_sum(tps_image, -1), 3*0.95), tf.float32)

[image, prod_image, coarse_image, tps_image, mask_output, tps_mask] = sess.run(

[image, prod_image, coarse_image, tps_image, mask_output, tps_mask])

return image, prod_image, coarse_image, tps_image, mask_output, tps_mask

def main(unused_argv):

try:

os.mkdir(FLAGS.result_dir)

except:

pass

try:

os.mkdir(FLAGS.result_dir + "/images/")

except:

pass

batch_size = 1

# Feed into the refine module

image_holder = tf.placeholder(tf.float32, shape=[batch_size,256,192,3])

prod_image_holder = tf.placeholder(tf.float32, shape=[batch_size,256,192,3])

prod_mask_holder = tf.placeholder(tf.float32, shape=[batch_size,256,192,1])

coarse_image_holder = tf.placeholder(tf.float32, shape=[batch_size,256,192,3])

tps_image_holder = tf.placeholder(tf.float32, shape=[batch_size,256,192,3])

with tf.variable_scope("refine_generator") as scope:

select_mask = create_refine_generator(tps_image_holder,

coarse_image_holder)

select_mask = select_mask * prod_mask_holder

model_image_outputs = (select_mask * tps_image_holder +

(1 - select_mask) * coarse_image_holder)

saver = tf.train.Saver(var_list=[var for var in tf.trainable_variables()

if var.name.startswith("refine_generator")])

with tf.Session() as sess:

print("loading model from checkpoint")

checkpoint = tf.train.latest_checkpoint(FLAGS.checkpoint)

if checkpoint == None:

checkpoint = FLAGS.checkpoint

print(checkpoint)

saver.restore(sess, checkpoint)

# reading input data

test_info = open(FLAGS.test_label).read().splitlines()

for i in range(FLAGS.begin, FLAGS.end, batch_size):

# loading batch data

print(i)

images = np.zeros((batch_size,256,192,3))

prod_images = np.zeros((batch_size,256,192,3))

coarse_images = np.zeros((batch_size,256,192,3))

tps_images = np.zeros((batch_size,256,192,3))

mask_outputs = np.zeros((batch_size,256,192,1))

image_names = []

product_image_names = []

for j in range(i, i + batch_size):

info = test_info[j].split()

print(info)

image_name = info[0]

product_image_name = info[1]

image_names.append(image_name)

product_image_names.append(product_image_name)

try:

(image, prod_image, coarse_image,

tps_image, mask_output, tps_mask) = _process_image(image_name,

product_image_name, sess)

except:

continue

images[j-i] = image

prod_images[j-i] = prod_image

coarse_images[j-i] = coarse_image

tps_images[j-i] = tps_image

mask_outputs[j-i] = np.expand_dims(mask_output, -1)

# inference

feed_dict = {

image_holder: images,

prod_image_holder: prod_images,

coarse_image_holder: coarse_images,

tps_image_holder: tps_images,

prod_mask_holder: mask_outputs,

}

[image_output, sel_mask] = sess.run([model_image_outputs, select_mask],

feed_dict=feed_dict)

# write results

for j in range(batch_size):

step = 0