-
Notifications
You must be signed in to change notification settings - Fork 80
/
main_test_drunet.py
210 lines (170 loc) · 7.07 KB
/
main_test_drunet.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
import os.path
import math
import argparse
import time
import random
import numpy as np
from collections import OrderedDict
import logging
from torch.utils.data import DataLoader
from torch.utils.data.distributed import DistributedSampler
import torch
from utils import utils_logger
from utils import utils_image as util
from utils import utils_option as option
from utils.utils_dist import get_dist_info, init_dist
from data.select_dataset import define_Dataset
from models.select_model import define_Model
# OCR metrics
# First, must install Tesseract: https://tesseract-ocr.github.io/tessdoc/Installation.html
# Second, install CER/WER and tesseract python wrapper libraries
# pip install fastwer
# pip install pybind11
# pip install pytesseract
import pytesseract
import fastwer
def calculate_cer_wer(img_E, img_H):
# Transcribe ground-truth image to text
text_H = pytesseract.image_to_string(img_H).strip().replace('\n',' ')
# Transcribe estimated image to text
text_E = pytesseract.image_to_string(img_E).strip().replace('\n',' ')
cer = fastwer.score_sent(text_E, text_H, char_level=True)
wer = fastwer.score_sent(text_E, text_H)
return cer, wer
'''
# --------------------------------------------
# Testing code for DRUNet
# --------------------------------------------
# Kai Zhang ([email protected])
# github: https://github.com/cszn/KAIR
# --------------------------------------------
# Adapted by Emilio Martínez ([email protected])
'''
def main(json_path='options/test_drunet.json'):
'''
# ----------------------------------------
# Step--1 (prepare opt)
# ----------------------------------------
'''
parser = argparse.ArgumentParser()
parser.add_argument('--opt', type=str, default=json_path, help='Path to option JSON file.')
parser.add_argument('--launcher', default='pytorch', help='job launcher')
parser.add_argument('--local_rank', type=int, default=0)
parser.add_argument('--dist', default=False)
opt = option.parse(parser.parse_args().opt, is_train=True)
opt['dist'] = parser.parse_args().dist
# ----------------------------------------
# configure logger
# ----------------------------------------
logger_name = 'test'
utils_logger.logger_info(logger_name, os.path.join(opt['path']['log'], logger_name + '.log'))
logger = logging.getLogger(logger_name)
logger.info(option.dict2str(opt))
# ----------------------------------------
# distributed settings
# ----------------------------------------
if opt['dist']:
init_dist('pytorch')
opt['rank'], opt['world_size'] = get_dist_info()
opt = option.dict_to_nonedict(opt)
model_path = opt['path']['pretrained_netG']
model_epoch = (model_path.split('/')[-1]).split('_G')[0]
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
opt_netG = opt['netG']
in_nc = opt_netG['in_nc']
out_nc = opt_netG['out_nc']
nc = opt_netG['nc']
nb = opt_netG['nb']
act_mode = opt_netG['act_mode']
bias = opt_netG['bias']
from models.network_unet import UNetRes as net
model = net(in_nc=in_nc, out_nc=out_nc, nc=nc, nb=nb, act_mode=act_mode, bias=bias)
model.load_state_dict(torch.load(model_path), strict=True)
model.eval()
for k, v in model.named_parameters():
v.requires_grad = False
model = model.to(device)
logger.info('Model path: {:s}'.format(model_path))
number_parameters = sum(map(lambda x: x.numel(), model.parameters()))
logger.info('Params number: {}'.format(number_parameters))
"""
# ----------------------------------------
# Step--3 (load paths)
# ----------------------------------------
"""
L_paths = util.get_image_paths(opt['datasets']['test']['dataroot_L'])
H_paths = util.get_image_paths(opt['datasets']['test']['dataroot_H'])
noise_sigma = opt['datasets']['test']['sigma_test']
'''
# ----------------------------------------
# Step--4 (main test)
# ----------------------------------------
'''
avg_psnr = 0.0
avg_ssim = 0.0
avg_edgeJaccard = 0.0
avg_cer = 0.0
avg_wer = 0.0
idx = 0
for L_path, H_path in zip(L_paths,H_paths):
idx += 1
image_name_ext = os.path.basename(L_path)
img_name, ext = os.path.splitext(image_name_ext)
img_dir = os.path.join(opt['path']['images'], img_name)
util.mkdir(img_dir)
logger.info('Creating inference on test image...')
# Load image
img_L_original = util.imread_uint(L_path, n_channels=3)[50:-50,100:-100,:]
img_L = img_L_original[:,:,:2]
img_L = util.uint2single(img_L)
img_L = util.single2tensor4(img_L)
# Add noise
if noise_sigma > 0:
noise_level = torch.FloatTensor([int(noise_sigma)])/255.0
noise = torch.randn(img_L.size()).mul_(noise_level).float()
img_L.add_(noise)
img_L = img_L.to(device)
# Inference on image
img_E = model(img_L)
img_L_tmp = util.tensor2uint(img_L)
img_L = np.zeros_like(img_L_original)
img_L[:,:,:2] = img_L_tmp
img_E = util.tensor2uint(img_E)
# -----------------------
# save noisy L
# -----------------------
save_img_path = os.path.join(img_dir, '{:s}_{}std.png'.format(img_name, noise_sigma))
util.imsave(img_L, save_img_path)
# -----------------------
# save estimated image E
# -----------------------
save_img_path = os.path.join(img_dir, '{:s}_model{}_{}std.png'.format(img_name, model_epoch, noise_sigma))
util.imsave(img_E, save_img_path)
logger.info(f'Inference of {img_name} completed. Saved at {img_dir}.')
# Load H image and compute metrics
img_H = util.imread_uint(H_path, n_channels=3)
if img_H.ndim == 3:
img_H = np.mean(img_H, axis=2)
img_H = img_H.astype('uint8')
# ----------------------------------------
# compute PSNR, SSIM, edgeJaccard and CER
# ----------------------------------------
current_psnr = util.calculate_psnr(img_E, img_H)
current_ssim = util.calculate_ssim(img_E, img_H)
current_edgeJaccard = util.calculate_edge_jaccard(img_E, img_H)
current_cer, current_wer = calculate_cer_wer(img_E, img_H)
logger.info('{:->4d}--> {:>10s} | PSNR = {:<4.2f}dB ; SSIM = {:.3f} ; edgeJaccard = {:.3f} ; CER = {:.3f}% ; WER = {:.3f}%'.format(idx, image_name_ext, current_psnr, current_ssim, current_edgeJaccard, current_cer, current_wer))
avg_psnr += current_psnr
avg_ssim += current_ssim
avg_edgeJaccard += current_edgeJaccard
avg_cer += current_cer
avg_wer += current_wer
avg_psnr = avg_psnr / idx
avg_ssim = avg_ssim / idx
avg_edgeJaccard = avg_edgeJaccard / idx
avg_cer = avg_cer / idx
avg_wer = avg_wer / idx
# Average log
logger.info('[Average metrics] PSNR : {:<4.2f}dB, SSIM = {:.3f} : edgeJaccard = {:.3f} : CER = {:.3f}% : WER = {:.3f}%'.format(avg_psnr, avg_ssim, avg_edgeJaccard, avg_cer, avg_wer))
if __name__ == '__main__':
main()