-
Notifications
You must be signed in to change notification settings - Fork 178
/
bts_test.py
220 lines (183 loc) · 9.5 KB
/
bts_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
215
216
217
218
219
220
# Copyright (C) 2019 Jin Han Lee
#
# This file is a part of BTS.
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>
from __future__ import absolute_import, division, print_function
import os
import argparse
import time
import numpy as np
import cv2
import sys
import torch
import torch.nn as nn
from torch.autograd import Variable
from bts_dataloader import *
import errno
import matplotlib.pyplot as plt
from tqdm import tqdm
from bts_dataloader import *
def convert_arg_line_to_args(arg_line):
for arg in arg_line.split():
if not arg.strip():
continue
yield arg
parser = argparse.ArgumentParser(description='BTS PyTorch implementation.', fromfile_prefix_chars='@')
parser.convert_arg_line_to_args = convert_arg_line_to_args
parser.add_argument('--model_name', type=str, help='model name', default='bts_nyu_v2')
parser.add_argument('--encoder', type=str, help='type of encoder, vgg or desenet121_bts or densenet161_bts',
default='densenet161_bts')
parser.add_argument('--data_path', type=str, help='path to the data', required=True)
parser.add_argument('--filenames_file', type=str, help='path to the filenames text file', required=True)
parser.add_argument('--input_height', type=int, help='input height', default=480)
parser.add_argument('--input_width', type=int, help='input width', default=640)
parser.add_argument('--max_depth', type=float, help='maximum depth in estimation', default=80)
parser.add_argument('--checkpoint_path', type=str, help='path to a specific checkpoint to load', default='')
parser.add_argument('--dataset', type=str, help='dataset to train on, make3d or nyudepthv2', default='nyu')
parser.add_argument('--do_kb_crop', help='if set, crop input images as kitti benchmark images', action='store_true')
parser.add_argument('--save_lpg', help='if set, save outputs from lpg layers', action='store_true')
parser.add_argument('--bts_size', type=int, help='initial num_filters in bts', default=512)
if sys.argv.__len__() == 2:
arg_filename_with_prefix = '@' + sys.argv[1]
args = parser.parse_args([arg_filename_with_prefix])
else:
args = parser.parse_args()
model_dir = os.path.dirname(args.checkpoint_path)
sys.path.append(model_dir)
for key, val in vars(__import__(args.model_name)).items():
if key.startswith('__') and key.endswith('__'):
continue
vars()[key] = val
def get_num_lines(file_path):
f = open(file_path, 'r')
lines = f.readlines()
f.close()
return len(lines)
def test(params):
"""Test function."""
args.mode = 'test'
dataloader = BtsDataLoader(args, 'test')
model = BtsModel(params=args)
model = torch.nn.DataParallel(model)
checkpoint = torch.load(args.checkpoint_path)
model.load_state_dict(checkpoint['model'])
model.eval()
model.cuda()
num_params = sum([np.prod(p.size()) for p in model.parameters()])
print("Total number of parameters: {}".format(num_params))
num_test_samples = get_num_lines(args.filenames_file)
with open(args.filenames_file) as f:
lines = f.readlines()
print('now testing {} files with {}'.format(num_test_samples, args.checkpoint_path))
pred_depths = []
pred_8x8s = []
pred_4x4s = []
pred_2x2s = []
pred_1x1s = []
start_time = time.time()
with torch.no_grad():
for _, sample in enumerate(tqdm(dataloader.data)):
image = Variable(sample['image'].cuda())
focal = Variable(sample['focal'].cuda())
# Predict
lpg8x8, lpg4x4, lpg2x2, reduc1x1, depth_est = model(image, focal)
pred_depths.append(depth_est.cpu().numpy().squeeze())
pred_8x8s.append(lpg8x8[0].cpu().numpy().squeeze())
pred_4x4s.append(lpg4x4[0].cpu().numpy().squeeze())
pred_2x2s.append(lpg2x2[0].cpu().numpy().squeeze())
pred_1x1s.append(reduc1x1[0].cpu().numpy().squeeze())
elapsed_time = time.time() - start_time
print('Elapesed time: %s' % str(elapsed_time))
print('Done.')
save_name = 'result_' + args.model_name
print('Saving result pngs..')
if not os.path.exists(os.path.dirname(save_name)):
try:
os.mkdir(save_name)
os.mkdir(save_name + '/raw')
os.mkdir(save_name + '/cmap')
os.mkdir(save_name + '/rgb')
os.mkdir(save_name + '/gt')
except OSError as e:
if e.errno != errno.EEXIST:
raise
for s in tqdm(range(num_test_samples)):
if args.dataset == 'kitti':
date_drive = lines[s].split('/')[1]
filename_pred_png = save_name + '/raw/' + date_drive + '_' + lines[s].split()[0].split('/')[-1].replace(
'.jpg', '.png')
filename_cmap_png = save_name + '/cmap/' + date_drive + '_' + lines[s].split()[0].split('/')[
-1].replace('.jpg', '.png')
filename_image_png = save_name + '/rgb/' + date_drive + '_' + lines[s].split()[0].split('/')[-1]
elif args.dataset == 'kitti_benchmark':
filename_pred_png = save_name + '/raw/' + lines[s].split()[0].split('/')[-1].replace('.jpg', '.png')
filename_cmap_png = save_name + '/cmap/' + lines[s].split()[0].split('/')[-1].replace('.jpg', '.png')
filename_image_png = save_name + '/rgb/' + lines[s].split()[0].split('/')[-1]
else:
scene_name = lines[s].split()[0].split('/')[0]
filename_pred_png = save_name + '/raw/' + scene_name + '_' + lines[s].split()[0].split('/')[1].replace(
'.jpg', '.png')
filename_cmap_png = save_name + '/cmap/' + scene_name + '_' + lines[s].split()[0].split('/')[1].replace(
'.jpg', '.png')
filename_gt_png = save_name + '/gt/' + scene_name + '_' + lines[s].split()[0].split('/')[1].replace(
'.jpg', '.png')
filename_image_png = save_name + '/rgb/' + scene_name + '_' + lines[s].split()[0].split('/')[1]
rgb_path = os.path.join(args.data_path, './' + lines[s].split()[0])
image = cv2.imread(rgb_path)
if args.dataset == 'nyu':
gt_path = os.path.join(args.data_path, './' + lines[s].split()[1])
gt = cv2.imread(gt_path, -1).astype(np.float32) / 1000.0 # Visualization purpose only
gt[gt == 0] = np.amax(gt)
pred_depth = pred_depths[s]
pred_8x8 = pred_8x8s[s]
pred_4x4 = pred_4x4s[s]
pred_2x2 = pred_2x2s[s]
pred_1x1 = pred_1x1s[s]
if args.dataset == 'kitti' or args.dataset == 'kitti_benchmark':
pred_depth_scaled = pred_depth * 256.0
else:
pred_depth_scaled = pred_depth * 1000.0
pred_depth_scaled = pred_depth_scaled.astype(np.uint16)
cv2.imwrite(filename_pred_png, pred_depth_scaled, [cv2.IMWRITE_PNG_COMPRESSION, 0])
if args.save_lpg:
cv2.imwrite(filename_image_png, image[10:-1 - 9, 10:-1 - 9, :])
if args.dataset == 'nyu':
plt.imsave(filename_gt_png, np.log10(gt[10:-1 - 9, 10:-1 - 9]), cmap='Greys')
pred_depth_cropped = pred_depth[10:-1 - 9, 10:-1 - 9]
plt.imsave(filename_cmap_png, np.log10(pred_depth_cropped), cmap='Greys')
pred_8x8_cropped = pred_8x8[10:-1 - 9, 10:-1 - 9]
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_8x8.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_8x8_cropped), cmap='Greys')
pred_4x4_cropped = pred_4x4[10:-1 - 9, 10:-1 - 9]
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_4x4.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_4x4_cropped), cmap='Greys')
pred_2x2_cropped = pred_2x2[10:-1 - 9, 10:-1 - 9]
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_2x2.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_2x2_cropped), cmap='Greys')
pred_1x1_cropped = pred_1x1[10:-1 - 9, 10:-1 - 9]
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_1x1.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_1x1_cropped), cmap='Greys')
else:
plt.imsave(filename_cmap_png, np.log10(pred_depth), cmap='Greys')
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_8x8.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_8x8), cmap='Greys')
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_4x4.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_4x4), cmap='Greys')
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_2x2.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_2x2), cmap='Greys')
filename_lpg_cmap_png = filename_cmap_png.replace('.png', '_1x1.png')
plt.imsave(filename_lpg_cmap_png, np.log10(pred_1x1), cmap='Greys')
return
if __name__ == '__main__':
test(args)