forked from Chan-man00/fogeye
-
Notifications
You must be signed in to change notification settings - Fork 0
/
plot_epoch_progress.py
189 lines (132 loc) · 8.34 KB
/
plot_epoch_progress.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
import os
import subprocess
import argparse
import matplotlib.pyplot as plt
import gif
from ssim import SSIM
from ssim.utils import get_gaussian_kernel
from PIL import Image
from utils_stereofog import calculate_model_results, generate_stats_from_log
from general_imports import *
parser = argparse.ArgumentParser()
parser.add_argument('--checkpoints_path', required=True, help='path to images (should have subfolders trainA, trainB, valA, valB, etc)')
parser.add_argument('--model_name', type=str, default=None, help='Override model name (if it should not be inferred from the dataroot))')
parser.add_argument('--verbosity', type=int, default=1, help='verbosity level (i.e., print every `verbosity` epochs)')
parser.add_argument('--ratio', type=float, default=4/3, help='aspect ratio of the images (to undo the transformation from pix2pix model)')
parser.add_argument('--image_index', type=int, default=0, help='try another image if the first one is not suitable')
parser.add_argument('--output_type', type=str, default='image', help='output type (image or gif)')
args = parser.parse_args()
# dataroot = args.dataroot
checkpoints_path = args.checkpoints_path
if args.model_name is None:
model_name = checkpoints_path.split('/')[-1]
else:
model_name = args.model_name
verbosity = args.verbosity
ratio = args.ratio
image_index = args.image_index
output_type = args.output_type
# Function for producing each of the frame of the gif (documentation: https://pypi.org/project/gif/)
@gif.frame
def produce_fig(epoch_index, image_index):
epoch = epochs[::verbosity][epoch_index]
subpath = os.path.join(results_path, f'{epoch}', model_name, f'test_{epoch}', 'images')
try:
file = sorted([item for item in os.listdir(subpath) if 'fake' in item])[image_index]
except:
print(f'Image index {image_index} not found. Using default image index 0 instead.')
file = sorted([item for item in os.listdir(subpath) if 'fake' in item])[0]
image_index = 0 # make sure that all images use this index now
img = plt.imread(os.path.join(subpath, file))
fig, ax = plt.subplots(1, 5, figsize=(5*width_per_image, height_per_image))
ax[0].plot(epochs[::verbosity], mean_CW_SSIM_scores)
ax[0].scatter(epochs[::verbosity][epoch_index], mean_CW_SSIM_scores[epoch_index], color='r')
ax[0].set_xlabel('epoch')
ax[0].set_ylabel('CW-SSIM score')
ax[1].imshow(img, aspect='auto')
ax[1].axis('off')
ax[1].text(0., 0., f'epoch {epoch}', color='k', backgroundcolor='w', horizontalalignment='left', verticalalignment='bottom', transform=ax[1].transAxes)
ax[1].text(0.5, 1, 'fake', color='k', backgroundcolor='w', horizontalalignment='center', verticalalignment='top', transform=ax[1].transAxes)
ax[2].imshow(original_img_A, aspect='auto')
ax[2].axis('off')
ax[2].text(0.5, 1, 'clear real', color='k', backgroundcolor='w', horizontalalignment='center', verticalalignment='top', transform=ax[2].transAxes)
ax[3].imshow(original_img_B, aspect='auto')
ax[3].axis('off')
ax[3].text(0.5, 1, 'foggy real', color='k', backgroundcolor='w', horizontalalignment='center', verticalalignment='top', transform=ax[3].transAxes)
_, ax[4] = generate_stats_from_log(checkpoints_path, fig=fig, ax=ax[4], highlight_epoch=epoch)
plt.subplots_adjust(hspace=0, wspace=0)
plt.tight_layout()
return fig
# model_name = 'stereofog_pix2pix'
# dataroot = 'datasets/stereofog_images'
# checkpoints_directory = f'checkpoints/{model_name}'
# checkpoints_directory = dataroot.replace('datasets', 'checkpoints')
# Extracting the epochs from the checkpoints directory
epochs = sorted(list(set([int(item.replace('_net_G.pth', '').replace('_net_D.pth', '')) for item in os.listdir(checkpoints_path) if '.pth' in item and 'latest' not in item])))
# results_path = f'results/epochs/'
results_path = f"results/epochs/{model_name}_epochs"
# dataset_image_path = os.path.join(dataroot, 'A', 'test')
# dataset_image_file = [item for item in os.listdir(dataset_image_path)][0]
# dataset_image = plt.imread(os.path.join(dataset_image_path, dataset_image_file))
# ratio = dataset_image.shape[1] / dataset_image.shape[0]
# Setting width and height for the plots
width_per_image = 4
height_per_image = width_per_image / ratio
# Reading in the original images (these will just be replotted again for every epoch and do not change)
original_subpath_A = os.path.join(results_path, f'{epochs[0]}', model_name, f'test_{epochs[0]}', 'images')
original_file_A = sorted([item for item in os.listdir(original_subpath_A) if 'real_B' in item])[image_index]
original_img_A = plt.imread(os.path.join(original_subpath_A, original_file_A))
original_subpath_B = os.path.join(results_path, f'{epochs[0]}', model_name, f'test_{epochs[0]}', 'images')
original_file_B = sorted([item for item in os.listdir(original_subpath_B) if 'real_A' in item])[image_index]
original_img_B = plt.imread(os.path.join(original_subpath_B, original_file_B))
# fig, ax = plt.subplots(len(epochs), 3, figsize=(3*width_per_image, len(epochs)*height_per_image))
# ax = ax.flatten()
# Evaluating the CW-SSIM score for each epoch
gaussian_kernel_sigma = 1.5
gaussian_kernel_width = 11
gaussian_kernel_1d = get_gaussian_kernel(gaussian_kernel_width, gaussian_kernel_sigma)
mean_CW_SSIM_scores = []
for i, epoch in enumerate(epochs[::verbosity]):
# subpath = os.path.join(results_path, f'{epoch}', model_name, f'test_{epoch}', 'images')
# try:
# file = [item for item in os.listdir(subpath) if 'fake' in item][image_index]
# except:
# print(f'Image index {image_index} not found. Using default image index 0 instead.')
# file = [item for item in os.listdir(subpath) if 'fake' in item][0]
# image_index = 0 # make sure that all images use this index now
mean_SSIM, mean_CW_SSIM = calculate_model_results(os.path.join(results_path, str(epoch)), epoch=epoch, epoch_test=True)
mean_CW_SSIM_scores.append(mean_CW_SSIM)
# checking if the save directory exists
if not os.path.exists(f"results/epochs/{model_name}_epochs_results"):
os.makedirs(f"results/epochs/{model_name}_epochs_results")
if output_type == 'image':
fig = plt.figure(figsize=(3*width_per_image,len(epochs[::verbosity])*height_per_image))
ax = [fig.add_subplot(len(epochs[::verbosity]),3,i+1) for i in range(len(epochs[::verbosity])*3)]
for i, epoch in enumerate(epochs[::verbosity]):
subpath = os.path.join(results_path, f'{epoch}', model_name, f'test_{epoch}', 'images')
try:
file = sorted([item for item in os.listdir(subpath) if 'fake' in item])[image_index]
except:
print(f'Image index {image_index} not found. Using default image index 0 instead.')
file = sorted([item for item in os.listdir(subpath) if 'fake' in item])[0]
image_index = 0 # make sure that all images use this index now
img = plt.imread(os.path.join(subpath, file))
ax[3*i].imshow(img, aspect='auto')
ax[3*i].axis('off')
ax[3*i].text(0.1, 0.02, f'epoch {epoch}', color='k', backgroundcolor='w', horizontalalignment='center', verticalalignment='center', transform=ax[3*i].transAxes)
ax[3*i+2].imshow(original_img_A, aspect='auto')
ax[3*i+2].axis('off')
ax[3*i+1].imshow(original_img_B, aspect='auto')
ax[3*i+1].axis('off')
plt.subplots_adjust(hspace=0, wspace=0)
# plt.savefig(f"results/epochs/{model_name}_epochs_results/{model_name}_epochs_results.png")
plt.show()
print(f"results/epochs/{model_name}_epochs_results/{model_name}_epochs_results.png saved successfully.")
plt.plot(epochs[::verbosity], mean_CW_SSIM_scores)
# plt.savefig(f"results/epochs/{model_name}_epochs_results/{model_name}_epochs_results_scores.png")
plt.show()
elif output_type == 'gif':
# fastgif.make_gif(produce_fig, len(epochs[::verbosity]), 'test.gif', show_progress=True, writer_kwargs={'image_index': image_index})
frames = [produce_fig(i, image_index) for i in range(len(epochs[::verbosity]))]
gif.save(frames, f"results/epochs/{model_name}_epochs_results/{model_name}_epochs_results.gif", duration=500)
print(f"results/epochs/{model_name}_epochs_results/{model_name}_epochs_results.gif saved successfully.")