-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathmetrics.py
259 lines (215 loc) · 11.5 KB
/
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
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
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
#
# Copyright (C) 2023, Inria
# GRAPHDECO research group, https://team.inria.fr/graphdeco
# All rights reserved.
#
# This software is free for non-commercial, research and evaluation use
# under the terms of the LICENSE.md file.
#
# For inquiries contact [email protected]
#
from pathlib import Path
import os
from PIL import Image
import torch
import torchvision.transforms.functional as tf
from utils.loss_utils import ssim
from lpipsPyTorch import lpips
import json
from tqdm import tqdm
from utils.image_utils import sse, psnr, masked_psnr
from argparse import ArgumentParser
import matplotlib.pyplot as plt
def readImages(renders_dir, gt_dir):
renders = []
gts = []
image_names = []
for fname in sorted(os.listdir(renders_dir)):
render = Image.open(renders_dir / fname)
gt = Image.open(gt_dir / fname)
renders.append(tf.to_tensor(render).unsqueeze(0)[:, :3, :, :].cuda())
gts.append(tf.to_tensor(gt).unsqueeze(0)[:, :3, :, :].cuda())
image_names.append(fname)
return renders, gts, image_names
def readMasks(masks_dir):
masks = []
image_names = []
for fname in sorted(os.listdir(masks_dir)):
mask = Image.open(masks_dir / fname)
masks.append(tf.to_tensor(mask).unsqueeze(0)[:, :3, :, :].cuda())
image_names.append(fname)
return masks, image_names
def evaluate(model_paths, mask_type, viz=False):
img_eval_dict = {}
full_dict = {}
per_view_dict = {}
full_dict_polytopeonly = {}
per_view_dict_polytopeonly = {}
for scene_dir in model_paths:
print("Scene:", scene_dir)
img_eval_dict[scene_dir] = {}
full_dict[scene_dir] = {}
per_view_dict[scene_dir] = {}
full_dict_polytopeonly[scene_dir] = {}
per_view_dict_polytopeonly[scene_dir] = {}
suffix=''
if mask_type=='seen_mask':
test_dir = Path(scene_dir) / "test_seen_masked"
suffix='_seen_masked'
elif mask_type=='eroded_seen_mask':
test_dir = Path(scene_dir) / "eroded_seen_mask" / "test_seen_masked"
suffix='_eroded_seen_masked'
elif mask_type=='mask':
test_dir = Path(scene_dir) / "test_masked"
suffix='_masked'
else:
test_dir = Path(scene_dir) / "test"
for method in os.listdir(test_dir):
if method!='ours_'+str(args.iterations):
continue
print("Method:", method)
if viz:
viz_path = scene_dir+f'/{method}_viz_{mask_type}/'
os.makedirs(viz_path, exist_ok=True)
img_eval_dict[scene_dir][method] = {}
full_dict[scene_dir][method] = {}
per_view_dict[scene_dir][method] = {}
full_dict_polytopeonly[scene_dir][method] = {}
per_view_dict_polytopeonly[scene_dir][method] = {}
method_dir = test_dir / method
gt_dir = method_dir/ "gt"
renders_dir = method_dir / "renders"
renders, gts, image_names = readImages(renders_dir, gt_dir)
valid_image_names = []
if mask_type=='seen_mask' or mask_type=='mask' or mask_type=='eroded_seen_mask':
masks_dir = method_dir / "masks"
masks, mask_image_names = readMasks(masks_dir)
if mask_type=='seen_mask' or mask_type=='mask' or mask_type=='eroded_seen_mask':
img_psnrs = []
img_ssims = []
img_lpipss = []
psnrs = []
sses = []
num_valid_pixels = []
for idx in tqdm(range(len(renders)), desc="Metric evaluation progress"):
# masked_psnr_ = masked_psnr(renders[idx], gts[idx], masks[idx])
# if masked_psnr_ is not None:
# psnrs.append(masked_psnr_)
sse_ = sse(renders[idx], gts[idx])
sses.append(sse_)
img_psnr_ = psnr(renders[idx], gts[idx])
img_ssim_ = ssim(renders[idx], gts[idx])
img_lpips_ = lpips(renders[idx], gts[idx])
num_valid_pixels_ = masks[idx].squeeze().sum()
if num_valid_pixels_ > masks[idx].squeeze().numel()/4.:
img_psnrs.append(img_psnr_)
img_ssims.append(img_ssim_)
img_lpipss.append(img_lpips_)
else:
img_psnr_ = torch.tensor(-1)
img_ssim_ = torch.tensor(-1)
img_lpips_ = torch.tensor(-1)
if num_valid_pixels_>0:
valid_image_names.append(image_names[idx])
num_valid_pixels.append(num_valid_pixels_)
psnr_ = 20 * torch.log10(1.0 / torch.sqrt(sse_/num_valid_pixels_+1e-5))
psnrs.append(psnr_)
if viz:
mse = (sse_/num_valid_pixels_).item()
mse = "{:.4f}".format(mse)
render_ = renders[idx].squeeze().permute(1,2,0).cpu().numpy()
gt_ = gts[idx].squeeze().permute(1,2,0).cpu().numpy()
mask_ = masks[idx].squeeze().permute(1,2,0).cpu().numpy()
err_img = (((renders[idx] - gts[idx])) ** 2).squeeze().permute((1,2,0))
# mask_out_idx = (mask_[:,:,0]==0)
# render_[:,:,0][mask_out_idx] = 0.62
# render_[:,:,1][mask_out_idx] = 0.13
# render_[:,:,2][mask_out_idx] = 0.94
# gt_[:,:,0][mask_out_idx] = 0.62
# gt_[:,:,1][mask_out_idx] = 0.13
# gt_[:,:,2][mask_out_idx] = 0.94
# err_img[:,:,0][mask_out_idx] = 0.62
# err_img[:,:,1][mask_out_idx] = 0.13
# err_img[:,:,2][mask_out_idx] = 0.94
fig, axs = plt.subplots(1, 3, figsize=(15, 5))
axs[0].imshow(render_)
axs[0].set_title(f'render\n mse: {mse}\n psnr:{img_psnr_.item()}\n ssim:{img_ssim_.item()}\n lpips:{img_lpips_.item()}')
axs[0].axis('off')
axs[1].imshow(gt_)
axs[1].set_title('gt')
axs[1].axis('off')
axs[2].imshow(err_img.cpu().detach().numpy())
axs[2].set_title('err_img')
axs[2].axis('off')
fig.tight_layout()
# plt.show()
plt.savefig(viz_path+image_names[idx])
plt.close()
print(" individual PSNR : {:>12.7f}".format(torch.tensor(psnrs).mean(), ".5"))
total_mse = torch.tensor(sses).sum() / torch.tensor(num_valid_pixels).sum()
total_psnr = 20 * torch.log10(1.0 / torch.sqrt(total_mse))
print(" total PSNR : {:>12.7f}".format(total_psnr, ".5"))
print("")
# full_dict[scene_dir][method].update({"PSNR": torch.tensor(psnrs).mean().item()})
full_dict[scene_dir][method].update({"PSNR": total_psnr.item()})
per_view_dict[scene_dir][method].update({"PSNR": {name: psnr for psnr, name in zip(torch.tensor(psnrs).tolist(), image_names)},
"Num Valid Pixels": {name: num_valid_pixels_ for num_valid_pixels_, name in zip(torch.tensor(num_valid_pixels).tolist(), image_names)}})
avg_img_psnr = torch.tensor(img_psnrs).mean()
avg_img_ssim = torch.tensor(img_ssims).mean()
avg_img_lpips = torch.tensor(img_lpipss).mean()
print("Image PSNR: ", avg_img_psnr.item())
print("Image SSIM: ", avg_img_ssim.item())
print("Image LPIPS: ", avg_img_lpips.item())
img_eval_dict[scene_dir][method].update({"PSNR": avg_img_psnr.item(),
"SSIM": avg_img_ssim.item(),
"LPIPS": avg_img_lpips.item()})
else:
ssims = []
psnrs = []
lpipss = []
for idx in tqdm(range(len(renders)), desc="Metric evaluation progress"):
ssims.append(ssim(renders[idx], gts[idx]))
psnrs.append(psnr(renders[idx], gts[idx]))
lpipss.append(lpips(renders[idx], gts[idx], net_type='vgg'))
print(" SSIM : {:>12.7f}".format(torch.tensor(ssims).mean(), ".5"))
print(" PSNR : {:>12.7f}".format(torch.tensor(psnrs).mean(), ".5"))
print(" LPIPS: {:>12.7f}".format(torch.tensor(lpipss).mean(), ".5"))
print("")
full_dict[scene_dir][method].update({"SSIM": torch.tensor(ssims).mean().item(),
"PSNR": torch.tensor(psnrs).mean().item(),
"LPIPS": torch.tensor(lpipss).mean().item()})
per_view_dict[scene_dir][method].update({"SSIM": {name: ssim for ssim, name in zip(torch.tensor(ssims).tolist(), image_names)},
"PSNR": {name: psnr for psnr, name in zip(torch.tensor(psnrs).tolist(), image_names)},
"LPIPS": {name: lp for lp, name in zip(torch.tensor(lpipss).tolist(), image_names)}})
with open(scene_dir + "/"+method+"_img_eval_results"+suffix+".json", 'w') as fp:
json.dump(img_eval_dict[scene_dir], fp, indent=True)
with open(scene_dir + "/"+method+"_results"+suffix+".json", 'w') as fp:
json.dump(full_dict[scene_dir], fp, indent=True)
with open(scene_dir + "/"+method+"_per_view"+suffix+".json", 'w') as fp:
json.dump(per_view_dict[scene_dir], fp, indent=True)
if __name__ == "__main__":
device = torch.device("cuda:0")
torch.cuda.set_device(device)
# Set up command line argument parser
parser = ArgumentParser(description="Training script parameters")
parser.add_argument('--model_paths', '-m', required=True, nargs="+", type=str, default=[])
parser.add_argument("--iterations", type=int, default=2000)
parser.add_argument('--no_mask', action="store_true")
parser.add_argument('--mask', action="store_true")
parser.add_argument('--seen_mask', action="store_true")
parser.add_argument('--eroded_seen_mask', action="store_true")
parser.add_argument('--viz', action="store_true")
args = parser.parse_args()
mask_types=[]
if args.no_mask:
mask_types.append("None")
if args.mask:
mask_types.append("mask")
if args.eroded_seen_mask:
mask_types.append("eroded_seen_mask")
if args.seen_mask:
mask_types.append("seen_mask")
if len(mask_types)==0:
print('Please specify the evaluation mode --[no_mask | mask | seen_mask]')
for mask_type in mask_types:
evaluate(args.model_paths, mask_type, args.viz)