AVMR/utils.py at master · yawenzeng/AVMR · GitHub

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
import numpy as np
import os
import glob
import torch as th
from torchvision import models, transforms
from PIL import Image
from torch.autograd import Variable

def dataloader(path):
    dataloader=np.load(path)
    data_all=list()
    for i in range(len(dataloader)):
        data = list()
        data.append(dataloader[i]['video_info'])
        data.append(dataloader[i]['w_vec'])
        data.append(dataloader[i]['v_feature'])
        data.append(dataloader[i]['w_start'])
        data.append(dataloader[i]['w_end'])
        data.append(dataloader[i]['fps'])

        data_all.append(data)

    return data_all

def extractor(img_path, net, use_gpu):
    transform = transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        transforms.ToTensor()]
    )
    img = Image.open(img_path)
    img = transform(img)

    x = Variable(th.unsqueeze(img, dim=0).float(), requires_grad=False)
    if use_gpu:
        x = x.cuda()
        net = net.cuda()
    y = net(x).cpu()
    y = y.data.numpy()

    return y.tolist()

def resneti(path,start,end):
    files_list=list()
    for i in range(start,end):
        file_glob = os.path.join(path, str(i) + '.jpg')
        files_list.extend(glob.glob(file_glob))

    resnet50_feature_extractor = models.resnet50(pretrained=False)
    resnet50_feature_extractor.load_state_dict(th.load('/resnet50-19c8e357.pth'))
    resnet50_feature_extractor.fc = th.nn.Linear(2048, 2048)
    th.nn.init.eye_(resnet50_feature_extractor.fc.weight)
    for param in resnet50_feature_extractor.parameters():
        param.requires_grad = False

    use_gpu = th.cuda.is_available()

    video_fearture=list()
    for x_path in [files_list[j] for j in range(len(files_list))]:
        sigle_feature = extractor(x_path, resnet50_feature_extractor, use_gpu)
        try:
            sigle_feature=extractor(x_path, resnet50_feature_extractor, use_gpu)
        except: break
        video_fearture.append(sigle_feature)
    feature=np.squeeze(np.array(video_fearture).mean(axis=0))

    return feature

def resnet(path,start,end):
    files_list=list()
    for i in range(start,end):
        file_glob = os.path.join(path, str(i) + '.jpg')
        files_list.extend(glob.glob(file_glob))

    resnet50_feature_extractor = models.resnet50(pretrained=False)
    resnet50_feature_extractor.load_state_dict(th.load('/resnet50-19c8e357.pth'))
    resnet50_feature_extractor.fc = th.nn.Linear(2048, 2048)
    th.nn.init.eye_(resnet50_feature_extractor.fc.weight)
    for param in resnet50_feature_extractor.parameters():
        param.requires_grad = False

    use_gpu = th.cuda.is_available()

    video_fearture=list()
    for x_path in [files_list[j] for j in range(len(files_list))]:
        sigle_feature = extractor(x_path, resnet50_feature_extractor, use_gpu)
        try:
            sigle_feature=extractor(x_path, resnet50_feature_extractor, use_gpu)
        except: break
        video_fearture.append(sigle_feature)

    feature=np.squeeze(video_fearture)
    return feature

def get_bpr(T,F):
    r=0
    if T>F: r=0.1
    elif T==F:r=0.5
    else: r=1
    return r

def calculate_IoU(i0, i1):
    # calculate temporal intersection over union
    union = (min(i0[0], i1[0]), max(i0[1], i1[1]))
    inter = (max(i0[0], i1[0]), min(i0[1], i1[1]))
    iou = 1.0*(abs(inter[1]-inter[0]))/(abs(union[1]-union[0]))
    return iou

def calculate_reward(Previou_IoU, current_IoU, t):
    if current_IoU > Previou_IoU and Previou_IoU>=0:
        reward = 1-0.00*t
    elif current_IoU <= Previou_IoU and current_IoU>=0:
        reward = -0.00*t
    else:
        reward = -1-0.00*t
    return reward

def compute_IoU_recall_top_n(topn, IoU, iuo_record):
    yes=0
    for i in range(len(iuo_record)):
        if max(iuo_record[i][:topn])>=IoU:
            yes=yes+1
    acc=yes/len(iuo_record)

    return acc