-
Notifications
You must be signed in to change notification settings - Fork 0
/
validate_CUHK.py
117 lines (86 loc) · 4.44 KB
/
validate_CUHK.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
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
"""
Created on Sun Aug 4 09:28:04 2019
@author: dragon
"""
import os
import tqdm
import torch as t
import numpy as np
import scipy.io as sio
from model.loss import l2_loss
from utils.config import config
from data.FlowData import FlowTestData
from torch.autograd import Variable
from model.networks import NetD, NetG
from torchnet.meter import AverageValueMeter
from utils.visualize import Visualizer
from sklearn.metrics import roc_auc_score
from sklearn.manifold import TSNE
from model_tv.network import model
from model_tv.train_options import arguments
if __name__ == '__main__':
args = arguments().parse()
args.data_size = [config.batch_size, config.iChanel, config.iWidth, config.iHeight]
netg, netd = NetG(config), NetD(config)
net_st_fusion = model(args).cuda()
error_meter = AverageValueMeter()
map_location=lambda storage, loc: storage
#netg.load_state_dict(t.load('cpkt_CUHK/netg_300.pth', map_location = map_location))
#netd.load_state_dict(t.load('cpkt_CUHK/netd_300.pth', map_location = map_location))
#net_st_fusion.load_state_dict(t.load('cpkt_CUHK/netfusion_300.pth', map_location = map_location))
if config.gpu:
netg.cuda()
netd.cuda()
net_st_fusion.cuda()
criterion_L2 = l2_loss
tsne=TSNE()
vis = Visualizer('Validation_CUHK')
root_path = 'dataset/Avenue_Dataset/testing_videos'
Video_Paths = os.listdir(root_path)
Video_Paths.sort()
Videos_objs = [os.path.join(video_path, video_path) for video_path in Video_Paths]
for iCnt in range(len(Videos_objs)):
video_dir = os.path.join(root_path, Video_Paths[iCnt])
datapair = FlowTestData(video_dir)
dataloader = t.utils.data.DataLoader(datapair,
batch_size = config.batch_size,
shuffle = False,
num_workers= config.num_workers,
drop_last=True
)
#latent_vector = np.zeros(int(datapair.imgs_len))
latent_vector1 = np.zeros(int(datapair.imgs_len))
error_meter.reset()
for ii,(videos) in tqdm.tqdm(enumerate(dataloader)):
img_3d = Variable(videos)
if config.gpu:
img_3d = img_3d.cuda()
img_st = net_st_fusion.forward(img_3d, need_result=True)
fake, latent_i, latent_o = netg.forward(img_st)
error = criterion_L2(latent_i, latent_o)
error_meter.add(error.data.cpu().numpy())
#latent_vector[ii] = error_meter.value()[0]
latent_vector1[ii] = error.data.cpu().numpy()
vis.plot('Normal_%s' %iCnt, error.data.cpu().numpy())
#vis.plot('Cardiac%s' %Video_Paths[iCnt], error.data.cpu().numpy())
#vis.images(Attention_img[0,:,16,:,:].detach().cpu().numpy()*127.0 + 127.0, win='Input')
#vis.images(fake[0,:,16,:,:].detach().cpu().numpy()*127.0 + 127.0, win='Reconstruct')
#vis.images((Attention_img[0,:,16,:,:] - fake[0,:,16,:,:]).detach().cpu().numpy()*127.0 + 127.0, win='Residual')
#s = Mask_list[0][0,0,16,:,:].detach().cpu().numpy().squeeze()
#s_m = np.rot90(s, 2)
#vis.heatmap(np.fliplr(s_m) , win='Mask')
#tsne.fit_transform(latent_i.detach().cpu().view(1, 29*29*100).numpy())
#latent_mat_path = os.path.join('result', 'Abnormal_%s.mat' %iCnt)
#sio.savemat(latent_mat_path, {'latent': latent_vector})
latent_mat_path = os.path.join('result/CUHK', 'CUHK_%s.mat' %Video_Paths[iCnt])
sio.savemat(latent_mat_path, {'latent': latent_vector1})
#vec_Normal = latent_vector
#latent_mat_path = os.path.join('result', 'abnormal_result.mat')
#sio.loadmat(latent_mat_path, {'latent': latent_vector})
#latent_vector = np.concatenate((latent_vector, vec_Normal), axis=0)
#latent_label1 = np.zeros(int(datapair.imgs_len))
#latent_label2 = np.ones(int(datapair.imgs_len))
#latent_label = np.concatenate((latent_label1, latent_label2), axis=0)
#print "AUC is ",roc_auc_score(latent_label, latent_vector)