/
thumos_i3d_per_video.py
141 lines (111 loc) · 4.02 KB
/
thumos_i3d_per_video.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
import torch
import torch.utils.data as data_utl
from torch.utils.data.dataloader import default_collate
import numpy as np
import json
import csv
import h5py
import os
import os.path
def video_to_tensor(pic):
"""Convert a ``numpy.ndarray`` to tensor.
Converts a numpy.ndarray (T x H x W x C)
to a torch.FloatTensor of shape (C x T x H x W)
Args:
pic (numpy.ndarray): Video to be converted to tensor.
Returns:
Tensor: Converted video.
"""
#return torch.from_numpy(pic.transpose([3, 0, 1, 2]))
return torch.from_numpy(pic.reshape((-1,1024)))
def make_dataset(split_file):
dataset = []
with open(split_file, 'r') as f:
data = json.load(f)
i = 0
for vid in data.keys():
gt = []
label = np.zeros(2, np.int64)
id = data[vid]['actions'][0][0]
label[0] = id # 从1开始
for ann in data[vid]['actions']:
ann = [ann[0],int(ann[1]*30),int(ann[2]*30)] # 转换为帧 30fps
gt.append(ann)
if label[0]!=int(ann[0]):
label[1]=int(ann[0])
dataset.append((vid, label, data[vid]['frame'], gt))
i += 1
return dataset
class Thumos(data_utl.Dataset):
def __init__(self, split_file, root, batch_size, mode, train):
self.data = make_dataset(split_file)
self.split_file = split_file
self.batch_size = batch_size
self.root = root
self.in_mem = {}
self.mode = mode
self.train = train
def __getitem__(self, index):
"""
Args:
index (int): Index
Returns:
tuple: (image, target) where target is class_index of the target class.
"""
entry = self.data[index] # name, label, frame, gt
feature=[]
if entry[0] in self.in_mem:
feat = self.in_mem[entry[0]]
else:
## concat rgb and flow featuremap
# for i in range(1,401):
# feat = np.load(os.path.join(self.root, entry[0], 'flow/'+ str(i) + '.npy'))
# feat_rgb = np.load(os.path.join(self.root, entry[0], 'rgb/' + str(i) + '.npy'))
# feat = feat.reshape((-1, 1024))
# feat_rgb = feat_rgb.reshape((-1, 1024))
# feat = feat.astype(np.float32)
# feat_rgb = feat_rgb.astype(np.float32)
# feat = np.concatenate((feat,feat_rgb),axis=1)
# feature.append(feat)
## 固定400段切分
# during = entry[2]/400
#
# for i in range(1,entry[2]):
# if i*during + 49 > entry[2]:
# continue
## 等间隔切分
during = 24
if entry[2] > 15000: # over 15000 frames
during = 48
if entry[2] > 30000:
during = 96
i=0
for item in range(0, entry[2] + 1, during):
if item + 49 > entry[2]:
continue
i+=1
feat = np.load(os.path.join(self.root, entry[0], self.mode +'/'+ str(i) + '.npy'))
feat = feat.astype(np.float32)
feature.append(feat)
feature = np.array(feature)
if self.train:
label = np.zeros(20) # for train multi-label
for i in range(len(entry[1])):
if entry[1][i]==0:
continue
label[entry[1][i]-1]=1
else:
label = entry[1]-1 # for test
return feature, label, [entry[0], entry[2], entry[3]]
def __len__(self):
return len(self.data)
def mt_collate_fn(batch):
# "Pads data and puts it into a tensor of same dimensions"
# max_len = 0
# for b in batch:
# if b[0].shape[0] > max_len:
# max_len = b[0].shape[0]
new_batch = []
for b in batch:
new_batch.append([video_to_tensor(b[0]),torch.from_numpy(b[1]) , b[2]])
return default_collate(new_batch)