/
test.py
214 lines (181 loc) · 8.19 KB
/
test.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
import os
import sys
import h5py
import numpy as np
import tensorflow.compat.v1 as tf
from loss import *
from data_generators import pyramid_inputs, load_sintel, load_sintel_fids
from data_generators import load_fcocc, load_fcocc_fids
from monet import monet
import matplotlib.pyplot as plt
from optflow import *
from scipy.io import savemat
from keras import optimizers
from argument_parser import myParser
def save_predictions(args, preds, preds2, fid, mbI=1, occI=1):
preds_occ = np.zeros((args.sizeV_orig, args.sizeH, 1))
preds_mb = np.zeros((args.sizeV_orig, args.sizeH, 1))
preds_occb = np.zeros((args.sizeV_orig, args.sizeH, 1))
preds_mbb = np.zeros((args.sizeV_orig, args.sizeH, 1))
mbInitI = 0 # 0-5
mbbInitI = 12 # 12-17
occInitI = 6 # 6-11
occbInitI = 18 # 18-23
if args.save_preds:
if not os.path.exists('predictions/'+args.experiment_name+
fid[:fid[1:].find('/')+1]):
os.mkdirs('predictions/'+args.experiment_name+
fid[:fid[1:].find('/')+1])
# mb predictions
if mbI == 1:
preds_mb[:args.sizeV_orig//2+1,:,0] \
= np.squeeze(preds[mbInitI+5])[:args.sizeV_orig//2+1,:]
preds_mb[-args.sizeV_orig//2:,:,0] \
= np.squeeze(preds2[mbInitI+5])[-args.sizeV_orig//2:,:]
preds_mbb[:args.sizeV_orig//2+1,:,0] \
= np.squeeze(preds[mbbInitI+5])[:args.sizeV_orig//2+1,:]
preds_mbb[-args.sizeV_orig//2:,:,0] \
= np.squeeze(preds2[mbbInitI+5])[-args.sizeV_orig//2:,:]
# save images
if args.save_preds:
plt.imsave('predictions/'+args.experiment_name+fid+'_mb.png',
np.squeeze(preds_mb), vmin=0.0, vmax=1.0)
np.save('predictions/'+args.experiment_name+fid+'_mb.npy',
np.squeeze(preds_mb))
plt.imsave('predictions/'+args.experiment_name+fid+'_mbb.png',
np.squeeze(preds_mbb), vmin=0.0, vmax=1.0)
np.save('predictions/'+args.experiment_name+fid+'_mbb.npy',
np.squeeze(preds_mbb))
# occ predictions
if occI == 1:
preds_occ[:args.sizeV_orig//2+1,:,0] \
= np.squeeze(preds[occInitI+5])[:args.sizeV_orig//2+1,:]
preds_occ[-args.sizeV_orig//2:,:,0] \
= np.squeeze(preds2[occInitI+5])[-args.sizeV_orig//2:,:]
preds_occb[:args.sizeV_orig//2+1,:,0] \
= np.squeeze(preds[occbInitI+5])[:args.sizeV_orig//2+1,:]
preds_occb[-args.sizeV_orig//2:,:,0] \
= np.squeeze(preds2[occbInitI+5])[-args.sizeV_orig//2:,:]
# save images
if args.save_preds:
plt.imsave('predictions/'+args.experiment_name+fid+'_occ.png',
np.squeeze(preds_occ), vmin=0.0, vmax=1.0)
np.save('predictions/'+args.experiment_name+fid+'_occ.npy',
np.squeeze(preds_occ))
plt.imsave('predictions/'+args.experiment_name+fid+'_occb.png',
np.squeeze(preds_occb), vmin=0.0, vmax=1.0)
np.save('predictions/'+args.experiment_name+fid+'_occb.npy',
np.squeeze(preds_occb))
return preds_mb, preds_mbb, preds_occ, preds_occb
def prepare_data(args, img1s, img2s, sps, sps2, flowEsts, bflowEsts):
img1 = img1s[:,:args.sizeV,:,:]
img2 = img2s[:,:args.sizeV,:,:]
sp = sps[:,:args.sizeV,:,:]
sp2 = sps2[:,:args.sizeV,:,:]
flowEst = flowEsts[:,:args.sizeV,:,:]
bflowEst = bflowEsts[:,:args.sizeV,:,:]
img1b = img1s[:,-args.sizeV:,:,:]
img2b = img2s[:,-args.sizeV:,:,:]
sp_2= sps[:,-args.sizeV:,:,:]
sp2_2 = sps2[:,-args.sizeV:,:,:]
flowEstb = flowEsts[:,-args.sizeV:,:,:]
bflowEstb = bflowEsts[:,-args.sizeV:,:,:]
# multi-scale inputs
img1s, sps, flowEsts = pyramid_inputs(img1, flowEst)
img2s, sps2, bflowEsts = pyramid_inputs(img2, bflowEst)
img1s_2, sps_2, flowEstsb = pyramid_inputs(img1b, flowEstb)
img2s_2, sps2_2, bflowEstsb = pyramid_inputs(img2b, bflowEstb)
X = [img1]+ img1s + [img2] + img2s + [sp] + sps + [sp2] + sps2 + \
[flowEst] + flowEsts + [bflowEst] + bflowEsts
X2 = [img1b]+ img1s_2 + [img2b] + img2s_2 + [sp_2] + sps_2 + \
[sp2_2] + sps2_2 + [flowEstb] + flowEstsb + [bflowEstb] + bflowEstsb
return X, X2
def main():
args = myParser()
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu_device
gpu_options = tf.GPUOptions(allow_growth=True,
per_process_gpu_memory_fraction=0.3)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options,
allow_soft_placement=True, log_device_placement=True))
if 'Sintel' in args.dataset_root:
print('Evaluating on MPI-Sintel dataset...')
fids = load_sintel_fids(args.dataset_root)
args.sizeV_orig = 436
args.sizeV = 416
args.sizeH = 1024
elif 'FlyingChairsOcc' in args.dataset_root:
print('Evaluating on FlyingChairsOcc dataset...')
fids = load_fcocc_fids(args.dataset_root)
args.sizeV_orig = 384
args.sizeV = 256
args.sizeH = 512
else:
print('Evaluating on a custom dataset...')
sys.exit('Need to setup dataloader for the custom dataset '
'(data_generators.py).')
# loading network
net = monet(args.sizeV, args.sizeH)
if args.load_weights is not None:
print ('Loading Network Weights: '+args.load_weights)
net.load_weights('experiments/'+args.load_weights, by_name=True)
# losses
mylosses, myweights = getMyJointLosses()
if args.optimizer_type == 'adam':
myO = optimizers.Adam(lr=args.learning_rate)
elif args.optimizer_type == 'sgd':
myO = optimizers.SGD(lr=args.learning_rate, decay=0.0001, momentum=0.9)
net.compile(loss=mylosses, loss_weights=myweights, optimizer=myO)
if 'Sintel' in args.dataset_root:
if 'final' in args.dataset_root:
finalI = True
args.experiment_name = args.experiment_name+'_final'
elif 'clean' in args.dataset_root:
finalI = False
args.experiment_name = args.experiment_name+'_clean'
elif 'FlyingChairsOcc' in args.dataset_root:
args.experiment_name = args.experiment_name+'_fcocc'
if args.save_preds:
print ('Saving images in '+'predictions/'+args.experiment_name)
if not os.path.exists('predictions/'+args.experiment_name):
os.mkdir('predictions/'+args.experiment_name)
beta=1
beta2 = beta ** 2
eps=1e-8
precision = []
recall = []
f1 = []
for i in range(len(fids)):
fid = fids[i]
# load inputs
if 'Sintel' in args.dataset_root:
img1s, img2s, sps, sps2, occ1, flowEst, bflowEst \
= load_sintel(args.dataset_root, args.flowEst_root,
fid=fids[i], finalI=finalI)
elif 'FlyingChairsOcc' in args.dataset_root:
img1s, img2s, sps, sps2, occ1, occ2, flowEst, bflowEst \
= load_fcocc(args.dataset_root, args.flowEst_root, fid=fids[i])
X, X2 = prepare_data(args, img1s, img2s, sps, sps2, flowEst, bflowEst)
# predict
preds = net.predict(X)
preds2 = net.predict(X2)
# save
preds_mb, preds_mbb, preds_occ, preds_occb \
= save_predictions(args, preds, preds2, fids[i])
# occlusion performance
preds_occ=np.squeeze(preds_occ>0.5).astype(float)
occ1=np.squeeze(occ1>0.5).astype(float)
true_positive=np.sum(preds_occ*occ1)
precision.append(true_positive/(np.sum(preds_occ)+eps))
recall.append(true_positive/(np.sum(occ1)+eps))
f1.append(precision[-1]*recall[-1]/(precision[-1]*beta2+recall[-1] \
+ eps) * (1 + beta2))
print(' ')
print('Occlusion Estimation:')
print(' Average precision: '+str(np.mean(precision)))
print(' Average recall: '+str(np.mean(recall)))
print(' Average F1: '+str(np.mean(f1)))
print(' ')
print('Motion Boundary Estimation:')
print(' Check README file for instructions.')
if __name__ == '__main__':
main()