This repository has been archived by the owner on May 18, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 31
/
Copy pathcalculate_metrics.py
144 lines (132 loc) · 7.64 KB
/
calculate_metrics.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
from skimage import data, img_as_float
from skimage.metrics import structural_similarity as ssim
from skimage.metrics import peak_signal_noise_ratio as psnr
from skimage.metrics import mean_squared_error
import argparse
from glob import glob
import os.path as osp
import os
import cv2
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from pathlib import Path
def argparser():
parser = argparse.ArgumentParser()
parser.add_argument(
"-t", "--ground_truth_dir_path", "--truth", type=str,
default="/home/gk32721/data/fw_gan_vvt/train/train_frames"
)
parser.add_argument(
"-g1", "--generated_dir_path_1", type=str,
default="/home/gk32721/2021-wacv-video-vton/result/VVTDataset/try-on"
)
parser.add_argument(
"-g2", "--generated_dir_path_2", type=str,
default="/home/gk32721/2021-wacv-video-vton/result/VVTDataset/try-on"
)
parser.add_argument(
"--experiment_one", type=str,
default="3.1"
)
parser.add_argument(
"--experiment_two", type=str,
default="3.2"
)
parser.add_argument(
"--result", type=str, default=""
)
return parser.parse_args()
def plot(args, df, video_folder):
df_frames = df.columns[1:]
indices = range(len(df_frames))
ssim_generated_1, ssim_generated_2 = df.loc["ssim_truth_generated_1"], df.loc["ssim_truth_generated_2"]
psnr_generated_1, psnr_generated_2 = df.loc["psnr_truth_generated_1"], df.loc["psnr_truth_generated_2"]
ssim_generated_1, ssim_generated_2 = ssim_generated_1.tolist()[1:], ssim_generated_2.tolist()[1:]
psnr_generated_1, psnr_generated_2 = psnr_generated_1.tolist()[1:], psnr_generated_2.tolist()[1:]
ssim_generated_1, ssim_generated_2 = [float(x) for x in ssim_generated_1], [float(x) for x in ssim_generated_2]
psnr_generated_1, psnr_generated_2 = [float(x) for x in psnr_generated_1], [float(x) for x in psnr_generated_2]
figs, axs = plt.subplots(2)
plt.suptitle(df.loc["file_path"][1].split('/')[0] + " Plot")
axs[0].plot(indices, ssim_generated_1, label="ssim_truth_generated_1", color="b")
axs[0].plot(indices, ssim_generated_2, label="ssim_truth_generated_2", color="g")
axs[1].plot(indices, psnr_generated_1, label="psnr_truth_generated_1", color="r")
axs[1].plot(indices, psnr_generated_2, label="psnr_truth_generated_2", color="y")
axs[0].legend()
axs[1].legend()
plt.xlabel("# of Frames")
plt.savefig(f"{args.result}/{args.experiment_one}_{args.experiment_two}_{video_folder}_plot.png")
plt.close()
def main():
args = argparser()
num_errors = 0
generated_image_frame_paths = glob(osp.join(args.generated_dir_path_1, '**', '*.png'))
metrics = {} # key: folder/file; value: (ssim, psnr)
assert len(generated_image_frame_paths) > 0
print(len(generated_image_frame_paths))
#from IPython import embed; embed()
#for root, folder, file in os.walk(args.generated_dir_path):
generated_video_folder_list_1 = sorted(os.listdir(args.generated_dir_path_1))
if not osp.exists(args.result):
os.mkdir(args.result)
for video_folder in generated_video_folder_list_1:
out = cv2.VideoWriter(f'{args.result}/compare_{args.experiment_one}_{args.experiment_two}_{video_folder}.mp4', cv2.VideoWriter_fourcc(*'DIVX'), 20.0, (768, 256))
generated_image_list = sorted(os.listdir(osp.join(args.generated_dir_path_1, video_folder)))
for i, image_file in enumerate(generated_image_list):
ground_truth_image_path = osp.join(args.ground_truth_dir_path, video_folder, image_file)
generated_image_frame_path_1 = osp.join(args.generated_dir_path_1, video_folder, image_file)
generated_image_frame_path_2 = osp.join(args.generated_dir_path_2, video_folder, image_file)
print(generated_image_frame_path_1)
print(generated_image_frame_path_2)
print(ground_truth_image_path)
assert osp.exists(generated_image_frame_path_1)
assert osp.exists(generated_image_frame_path_2)
assert osp.exists(ground_truth_image_path)
ground_truth_image = cv2.imread(ground_truth_image_path)
generated_image_1 = cv2.imread(generated_image_frame_path_1)
generated_image_2 = cv2.imread(generated_image_frame_path_2)
list_of_generated_images = [generated_image_1, generated_image_2]
details_numpy = np.zeros_like(generated_image_1)
#calculate metrics
print(type(generated_image_1))
print(type(ground_truth_image))
if generated_image_1 is None or generated_image_2 is None:
num_errors += 1
else:
ssim_metric = [ssim(ground_truth_image, generated_image,
data_range=generated_image.max() - generated_image.min(),
multichannel=True) for generated_image in list_of_generated_images]
psnr_metric = [psnr(ground_truth_image, generated_image,
data_range=generated_image.max() - generated_image.min())
for generated_image in list_of_generated_images]
# combine frames
combined_image = np.concatenate((ground_truth_image, list_of_generated_images[0], list_of_generated_images[1], details_numpy), axis=1)
#from IPython import embed;
#embed()
print(combined_image.shape)
cv2.putText(combined_image, f"Ground Truth Image", (20, 240), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (0, 0, 0))
cv2.putText(combined_image, f"Generated Image 1", (210, 240), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (0, 0, 0))
cv2.putText(combined_image, f"Generated Image 2", (450, 240), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (0, 0, 0))
cv2.putText(combined_image, f"SSIM_1 Metric: {ssim_metric[0]}", (600, 10), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (255,255,255))
cv2.putText(combined_image, f"SSIM_2 Metric: {ssim_metric[1]}", (600, 20), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (255, 255, 255))
cv2.putText(combined_image, f"PSNR_1 Metric: {psnr_metric[0]}", (600, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (255,255,255))
cv2.putText(combined_image, f"PSNR_2 Metric: {psnr_metric[1]}", (600, 40), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (255, 255, 255))
cv2.putText(combined_image, f"Frame: {i}", (600, 50), cv2.FONT_HERSHEY_SIMPLEX, 0.33, (255, 255, 255))
#cv2.imwrite('combined_image.jpg', combined_image)
cv2.cvtColor(combined_image, cv2.COLOR_RGB2BGR)
# add frames to video
out.write(combined_image)
# save in dictionary
metrics[i] = (osp.join(video_folder, image_file), ssim_metric[0], ssim_metric[1], psnr_metric[0], psnr_metric[1])
print("calculated")
print(ssim_metric)
print(psnr_metric)
out.release()
df = pd.DataFrame.from_dict(metrics)
df.index = ["file_path", "ssim_truth_generated_1", "ssim_truth_generated_2", "psnr_truth_generated_1", "psnr_truth_generated_2"]
plot(args, df, video_folder)
#print(f"{args.result}/compare_{args.experiment_one}_{args.experiment_two}_{video_folder}/{args.experiment_one}_{args.experiment_two}_metrics_output.csv")
#assert 1 == 0, os.getcwd()
df.to_csv(f"{args.result}/{args.experiment_one}_{args.experiment_two}_{video_folder}_metrics_output.csv")
if __name__ == '__main__':
main()