-
Notifications
You must be signed in to change notification settings - Fork 1
/
Testing_new.py
202 lines (188 loc) · 10.3 KB
/
Testing_new.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
import glob
import re
import os
import argparse
# from model_org import *
from model_org_raft_new_mod import *
import torch
import torch.optim as optim
from torch.utils.data import DataLoader
import json
import time
import PIL
from VTL_data import VTL_dataset
from torchvision import transforms
from PIL import Image
import scipy.io as sc
import logging
torch.backends.cudnn.enabled = True
gpu_num = torch.cuda.device_count()
feature_channel = 96
latent_channel = 192
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_sim_entr_0.125_noise_round_4/iter_32312.pth.tar'
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_sim_entr/deep_encoder_increased_latent_2048_temp_5_frames_sim_entr/iter_107730.pth.tar'
# pretrained_path = '//nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_sim_entr_noise_round_2_bit_0.25/iter_32312.pth.tar'
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_simp_entr_noise_round_0_bit_0.5/iter_55392.pth.tar'
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_comp_entr_noise_round_0_bit_0.5/iter_48468.pth.tar'
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_comp_entr_noise_round_0_bit_0.5_24_channel/iter_48468.pth.tar'
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_0.9_temp_5_frames_simp_entr_noise_round_0_bit_0.5/iter_55392.pth.tar'
# pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_comp_entr_noise_round_0_bit_0.5_16_channel/iter_46160.pth.tar'
pretrained_path = '/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_latent_2048_temp_5_frames_comp_entr_noise_round_0_bit_0.5_96_channel_sigmoid_v2/iter_50000.pth.tar'
root_dir = '/nfs/turbo/coe-hunseok/mrakeshc/VTL_dataset/'
# root_dir = '/nfs/turbo/coe-hunseok/mrakeshc/UVG_dataset/'
logger = logging.getLogger("UVG_testing")
def atoi(text):
return int(text) if text.isdigit() else text
def natural_keys(text):
'''
alist.sort(key=natural_keys) sorts in human order
http:https://nedbatchelder.com/blog/200712/human_sorting.html
(See Toothy's implementation in the comments)
'''
return [ atoi(c) for c in re.split(r'(\d+)', text) ]
def test():
sumBpp = 0
sumPsnr = 0
sumMsssim = 0
sumMsssimDB = 0
cnt = 0
residuals = []
root_dir_path = sorted(os.listdir(root_dir))
for dir_name in root_dir_path:
# for i in range(1):
# dir_name = root_dir_path[0]
image_list = []
folder_path = root_dir + dir_name + '/'
for filename in sorted(glob.iglob(folder_path + '**/*.png', recursive=True)):
image_list.append(filename)
image_list.sort(key=natural_keys)
video_len = len(image_list)
# print(image_list)
prev1 = PIL.Image.open(image_list[0]).convert("RGB")
prev2 = PIL.Image.open(image_list[1]).convert("RGB")
prev3 = PIL.Image.open(image_list[2]).convert("RGB")
prev4 = PIL.Image.open(image_list[3]).convert("RGB")
prev5 = PIL.Image.open(image_list[4]).convert("RGB")
torch_transform = transforms.Compose([
# transforms.CenterCrop((256,56)),
# transforms.CenterCrop((1024,1024)),
transforms.ToTensor(),
])
prev1 = torch_transform(prev1).unsqueeze(dim = 0)
prev2 = torch_transform(prev2).unsqueeze(dim = 0)
prev3 = torch_transform(prev3).unsqueeze(dim = 0)
prev4 = torch_transform(prev4).unsqueeze(dim = 0)
prev5 = torch_transform(prev5).unsqueeze(dim = 0)
with torch.no_grad():
num = 0
sum_per_BPP = 0
sum_per_psnr = 0
sum_per_mssim = 0
sum_per_mssimdb = 0
print("Currently the video compression is done for", dir_name)
net.eval()
incr = 0
# net.train()
for cur_idx in image_list[5:]:
incr += 1
dir_image = '/nfs/turbo/coe-hunseok/mrakeshc/Images_reconstructed_new_model/' + dir_name + 'VTL_' + str(incr)
cur = PIL.Image.open(cur_idx).convert("RGB")
cur = torch_transform(cur).unsqueeze(dim = 0)
# print(cur.shape)
clipped_recon_image, mse_loss, bpp_feature, bpp_z, bpp = net(prev1.to('cuda'), prev2.to('cuda'), prev3.to('cuda'), prev4.to('cuda'), prev5.to('cuda'), cur.to('cuda'), residuals, dir_image)
# clipped_recon_image, mse_loss, bpp_feature, bpp_z, bpp = net(prev1.to('cuda'), prev2.to('cuda'), prev3.to('cuda'), prev4.to('cuda'), cur.to('cuda'), residuals)
mse_loss, bpp_feature, bpp_z, bpp = \
torch.mean(mse_loss), torch.mean(bpp_feature), torch.mean(bpp_z), torch.mean(bpp)
psnr = 10 * (torch.log(1. / mse_loss) / np.log(10))
sumBpp += bpp
sumPsnr += psnr
cpu_recon_image = clipped_recon_image.cpu().detach()
np_recon_image = cpu_recon_image.permute(0,2,3,1).squeeze().numpy()
image = Image.fromarray((np_recon_image*255).astype(np.uint8))
save_path = '/nfs/turbo/coe-hunseok/mrakeshc/reconstructed_images_train/recon_VTL_sim_entr_new_model/' + dir_name + '/'
os.makedirs(save_path, exist_ok=True)
image.save(save_path + '/recon_{}.png'.format(num+1))
orig_image = cur.permute(0,2,3,1).squeeze().numpy()
orig_image = Image.fromarray((orig_image*255).astype(np.uint8))
orig_image.save(save_path + '/orig_{}.png'.format(num+1))
msssim = ms_ssim(cpu_recon_image, cur, data_range=1.0, size_average=True)
msssimDB = -10 * (torch.log(1-msssim) / np.log(10))
sumMsssimDB += msssimDB
sumMsssim += msssim
num += 1
cnt += 1
sum_per_BPP += bpp
sum_per_psnr += psnr
sum_per_mssim += msssim
sum_per_mssimdb += msssimDB
logger.info("Num: {}, Bpp:{:.6f}, PSNR:{:.6f}, MS-SSIM:{:.6f}, MS-SSIM-DB:{:.6f}".format(cnt, bpp, psnr, msssim, msssimDB))
print("Num: {}, Bpp:{:.6f}, PSNR:{:.6f}, MS-SSIM:{:.6f}, MS-SSIM-DB:{:.6f}".format(cnt, bpp, psnr, msssim, msssimDB))
prev1 = prev2
prev2 = prev3
prev3 = prev4
prev4 = clipped_recon_image
# if num == 20:
# break
# else:
# continue
# break
with open('Per_video_performance_VTLdeep_encoder_increased_latent_2048_temp_5_frames_comp_entr_noise_round_0_bit_0.5_96_channel_new.txt', 'a') as f_text:
f_text.write("Num: {}, Bpp:{:.6f}, PSNR:{:.6f}, MS-SSIM:{:.6f}, MS-SSIM-DB:{:.6f}".format(num, sum_per_BPP/num, sum_per_psnr/num, sum_per_mssim/num, sum_per_mssimdb/num))
f_text.write("\n")
# prev1 = prev2
# prev2 = clipped_recon_image
# prev1 = prev2
# prev2 = prev3
# prev3 = prev4
# prev4 = clipped_recon_image
# if num == 10:
# break
# else:
# continue
# break
# if cnt == 100:
# break
# print("Test on Vimeo triplet dataset: model-{}".format(step))
sumBpp /= cnt
sumPsnr /= cnt
sumMsssim /= cnt
sumMsssimDB /= cnt
logger.info("Dataset Average result---Dataset Num: {}, Bpp:{:.6f}, PSNR:{:.6f}, MS-SSIM:{:.6f}, MS-SSIM-DB:{:.6f}".format(cnt, sumBpp, sumPsnr, sumMsssim, sumMsssimDB))
print("Dataset Average result---Dataset Num: {}, Bpp:{:.6f}, PSNR:{:.6f}, MS-SSIM:{:.6f}, MS-SSIM-DB:{:.6f}".format(cnt, sumBpp, sumPsnr, sumMsssim, sumMsssimDB))
with open('Per_video_performance_VTLdeep_encoder_increased_latent_2048_temp_5_frames_comp_entr_noise_round_0_bit_0.5_96_channel_new.txt', 'a') as f_text:
f_text.write("\n")
f_text.write("Dataset Average result---Dataset Num: {}, Bpp:{:.6f}, PSNR:{:.6f}, MS-SSIM:{:.6f}, MS-SSIM-DB:{:.6f}".format(cnt, sumBpp, sumPsnr, sumMsssim, sumMsssimDB))
f_text.write("\n")
f_text.write("\n")
# sc.savemat('residuals_gt.mat', {'arr': residuals})
if __name__ == "__main__":
save_path = '/home/mrakeshc/RAFT_video_compression/'
formatter = logging.Formatter('%(asctime)s - %(levelname)s] %(message)s')
formatter = logging.Formatter('[%(asctime)s][%(filename)s][L%(lineno)d][%(levelname)s] %(message)s')
stdhandler = logging.StreamHandler()
stdhandler.setLevel(logging.ERROR)
stdhandler.setFormatter(formatter)
logger.addHandler(stdhandler)
tb_logger = None
filehandler = logging.FileHandler(os.path.join(save_path, 'log.txt'))
filehandler.setLevel(logging.INFO)
filehandler.setFormatter(formatter)
logger.addHandler(filehandler)
logger.info("Testing")
model = VideoCoder(feature_channel, latent_channel)
# filename_list = []
# for filename in glob.glob('/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_sim_entr_noise_round_2_bit_0.25/*.tar'): #assuming gif
# filename_list.append(filename)
# for filename in glob.glob('/nfs/turbo/coe-hunseok/mrakeshc/vimeo_triplet_video_compression/deep_encoder_increased_latent_2048_temp_5_frames_sim_entr_noise_round_2_bit_0.25/*/*.tar'): #assuming gif
# filename_list.append(filename)
# for i in range(len(filename_list)):
# pre_trained_path = filename_list[i]
# with open('Per_video_performance_VTL_sim_entr_noise_round_2_bit_0.25_multiple.txt', 'a') as f_text:
# f_text.write(pre_trained_path)
# f_text.write("\n")
# load_model(model, pre_trained_path)
# net = model.cuda()
# test()
load_model(model, pretrained_path)
net = model.cuda()
test()