-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmob_dataloader.py
250 lines (215 loc) · 9.62 KB
/
mob_dataloader.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
import os
import cv2
import random
import numpy as np
import skvideo.io
import pandas as pd
import os.path as osp
import torch
from torch.utils.data import DataLoader, Dataset
from torchvision import transforms
from torchvideotransforms.video_transforms import Compose, Resize, RandomCrop, RandomRotation, \
ColorJitter, Normalize, RandomHorizontalFlip, CenterCrop
from torchvideotransforms.volume_transforms import ClipToTensor
class MOBDataset(Dataset):
"""MOB dataset for recognition. The class index start from 0.
Args:
root_dir (string): Directory with videos and splits.
train (bool): train split or test split.
clip_len (int): number of frames in clip, 16/32/64.
transforms_ (object): composed transforms which takes in PIL image and output tensors.
test_sample_num： number of clips sampled from a video. 1 for clip accuracy.
"""
def __init__(self, root_dir, clip_len, split='1', train=True, transforms_=None, test_sample_num=1, train_file=None,
test_file=None, max_sr=1):
self.root_dir = root_dir
self.clip_len = clip_len
self.split = split
self.train = train
self.transforms_ = transforms_
self.test_sample_num = test_sample_num
self.toPIL = transforms.ToPILImage()
self.train_file = train_file
self.test_file = test_file
self.max_sr = max_sr
if self.train:
train_split_path = self.train_file
self.train_split = pd.read_csv(train_split_path, header=None, sep=' ')[0]
else:
test_split_path = self.test_file
self.test_split = pd.read_csv(test_split_path, header=None, sep=' ')[0]
def __len__(self):
if self.train:
return len(self.train_split)
else:
return len(self.test_split)
def __getitem__(self, idx):
#print('_'*100)
"""
Returns:
clip (tensor): [channel x time x height x width]
class_idx (tensor): class index, [0-100]
"""
if self.train:
videoname = self.train_split[idx]
else:
videoname = self.test_split[idx]
#print('parent_file',videoname)
class_idx = videoname.split('/')[-1]
vid_name = videoname.split('/')[1]# + '_' + videoname.split('_')[2]
folder_name = videoname.split('/')[0]
#print('folder_name',folder_name)
# exit()
#print(vid_name,class_idx)
#exit()
class_idx = int(class_idx) - 1
sample_rate = self.max_sr
filename = os.path.join(self.root_dir,folder_name,vid_name)
#print('filename',filename)
length = len(os.listdir(filename))
#print('length',length)
#exit()
try:
clip_start = random.randint(1, length - (self.clip_len * sample_rate))
except:
try:
clip_start = random.randint(1, length + 1 - (self.clip_len * sample_rate))
except:
clip_start = 1
clip = self.loop_load_rgb(filename, clip_start, sample_rate, self.clip_len, length)
#print('clip',clip.shape)
#exit()
sample = {}
# random select a clip for train
if self.train:
# clip_start = random.randint(0, length - self.clip_len)
# clip = videodata[clip_start: clip_start + self.clip_len] # uint8
# print(clip.shape, clip.dtype)
# exit()
if self.transforms_:
trans_clip_1 = []
trans_clip_2 = []
# fix seed, apply the sample `random transformation` for all frames in the clip
seed = random.random()
for frame in clip:
random.seed(seed)
framex = self.toPIL(frame) # PIL image
# framex = self.transforms_(framex) # tensor [C x H x W]
trans_clip_1.append(framex)
# (T x C X H x W) to (C X T x H x W)
# clip_1 = torch.stack(trans_clip_1).permute([1, 0, 2, 3])
clip_1 = self.transforms_(trans_clip_1)
else:
clip_1 = torch.tensor(clip)
#print(clip.shape, class_idx)
#exit()
sample['data_1'] = clip_1
sample['class_id'] = torch.tensor(class_idx)
return sample
# sample several clips for test
else:
# print(filename)
all_clips = []
all_idx = []
for i in np.linspace(self.clip_len / 2, length - self.clip_len, self.test_sample_num):
# clip_start = int(i - self.clip_len/2)
# clip = videodata[clip_start: clip_start + self.clip_len]
clip = list()
clip_start = int(i - self.clip_len / 2) + 1
clip_end = clip_start + self.clip_len
# print(clip_start, clip_end)
for ind_frame in range(clip_start, clip_end):
#print(osp.join(filename, str(ind_frame).zfill(4) + '.jpg'))
#exit()
# print(filename, str(ind_frame).zfill(5) + '.png')
# exit()
# cur_img_path = os.path.join(
# video_dir,
# # str(ind_frame).zfill(5) + '.png'
# "{:04}.jpg".format(start_frame + idx * sample_rate))
frm = cv2.cvtColor(cv2.imread(osp.join(filename, str(ind_frame).zfill(4) + '.jpg')),
cv2.COLOR_BGR2RGB)
# frm = cv2.resize(frm, (112, 112))
clip.append(frm)
# print(frm.shape)
clip = np.array(clip)
if self.transforms_:
trans_clip = []
# fix seed, apply the sample `random transformation` for all frames in the clip
seed = random.random()
for frame in clip:
random.seed(seed)
frame = self.toPIL(frame) # PIL image
# frame = self.transforms_(frame) # tensor [C x H x W]
trans_clip.append(frame)
# (T x C X H x W) to (C X T x H x W)
# clip = torch.stack(trans_clip).permute([1, 0, 2, 3])
clip = self.transforms_(trans_clip)
else:
clip = torch.tensor(clip)
all_clips.append(clip)
all_idx.append(torch.tensor(class_idx))
# print(all_idx, class_idx)
# sample['data'] = all_clips[0]
sample['data'] = all_clips
# sample['pid'] = all_idx
sample['class_id'] = torch.tensor(class_idx)
# print(sample['data'].shape, sample['pid'])
return sample
def loop_load_rgb(self, video_dir, start_frame, sample_rate, clip_len,
num_frames):
video_clip = []
idx = 0
for i in range(clip_len):
# cur_img_path = os.path.join(
# video_dir,
# # str(ind_frame).zfill(5) + '.png'
# "image_{:05}.png".format(start_frame + idx * sample_rate))
cur_img_path = os.path.join(
video_dir,
# str(ind_frame).zfill(5) + '.png'
"{:04}.jpg".format(start_frame + idx * sample_rate))
#print('cur_img_path',cur_img_path)
#exit()
img = cv2.cvtColor(cv2.imread(cur_img_path), cv2.COLOR_BGR2RGB)
# img = cv2.resize(img, (112, 112))
video_clip.append(img)
if (start_frame + (idx + 1) * sample_rate) > num_frames:
start_frame = 1
idx = 0
else:
idx += 1
video_clip = np.array(video_clip)
return video_clip
def build_dataloader():
train_transforms = Compose([
# RandomRotation(15),
Resize((256, 256)),
RandomCrop((224, 224)),
ColorJitter(0.5, 0.5, 0.5, 0.25),
RandomHorizontalFlip(),
ClipToTensor(),
Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
test_transforms = Compose([
# RandomRotation(15),
Resize((224, 224)),
ClipToTensor(),
Normalize(mean=(0.48145466, 0.4578275, 0.40821073), std=(0.26862954, 0.26130258, 0.27577711))
])
user_root = '/home/syedhammadahmed/Datasets' #todo: directory where /mob directory resides
mob_frames_path = '/mob/MOB_RUN/frames/video'
mob_frames_abs_path = user_root + mob_frames_path
train_path = './mob_dataloader/train_list.txt'
train_dataset = MOBDataset(mob_frames_abs_path,
8, train=True, transforms_=train_transforms, train_file=train_path)
test_path = './mob_dataloader/test_list.txt'
test_dataset = MOBDataset(mob_frames_abs_path,
8, train=False, transforms_=test_transforms, test_file=test_path)
train_dataloader = DataLoader(train_dataset,
batch_size=8, num_workers=4,
pin_memory=True, drop_last=True, shuffle=True)
test_dataloader = DataLoader(test_dataset,
batch_size=1, num_workers=4,
pin_memory=True, drop_last=True, shuffle=False)
return train_dataloader, test_dataloader, test_dataset