-
Notifications
You must be signed in to change notification settings - Fork 8
/
dataset.py
269 lines (239 loc) · 13.9 KB
/
dataset.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
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
# -*- coding: utf-8 -*-
import numpy as np
import pandas as pd
import json
import torch.utils.data as data
import torch
from utils import ioa_with_anchors, iou_with_anchors
from ipdb import set_trace
def load_json(file):
with open(file) as json_file:
json_data = json.load(json_file)
return json_data
class VideoDataSet(data.Dataset):
def __init__(self, opt, subset="train"):
self.temporal_scale = opt["temporal_scale"] # 100
self.temporal_gap = 1. / self.temporal_scale
self.subset = subset
self.mode = opt["mode"]
self.feature_path = opt["feature_path"]
self.video_info_path = "./data/activitynet_annotations/video_info_new_{}.csv".format(opt['unlabel_percent'])
self.video_anno_path = opt["video_anno"]
self._getDatasetDict()
self._get_match_map()
# set_trace()
def _getDatasetDict(self):
anno_df = pd.read_csv(self.video_info_path)
anno_database = load_json(self.video_anno_path)
self.video_dict = {}
for i in range(len(anno_df)):
video_name = anno_df.video.values[i]
video_info = anno_database[video_name]
video_subset = anno_df.subset.values[i]
if self.subset in video_subset:
if 'unlabel' not in video_subset:
self.video_dict[video_name] = video_info
self.video_list = list(self.video_dict.keys())
print("%s subset video numbers: %d" % (self.subset, len(self.video_list)))
def __getitem__(self, index):
video_data = self._load_file(index)
if self.mode == "train":
match_score_start, match_score_end, confidence_score = self._get_train_label(index, self.anchor_xmin,
self.anchor_xmax)
return video_data,confidence_score, match_score_start, match_score_end # [400,100],[100,100],[100]
else:
return index, video_data
def _get_match_map(self):
match_map = []
for idx in range(self.temporal_scale):
tmp_match_window = []
xmin = self.temporal_gap * idx
for jdx in range(1, self.temporal_scale + 1):
xmax = xmin + self.temporal_gap * jdx
tmp_match_window.append([xmin, xmax])
match_map.append(tmp_match_window)
match_map = np.array(match_map) # 100x100x2
match_map = np.transpose(match_map, [1, 0, 2]) # [0,1] [1,2] [2,3].....[99,100]
match_map = np.reshape(match_map, [-1, 2]) # [0,2] [1,3] [2,4].....[99,101] # duration x start
self.match_map = match_map # duration is same in row, start is same in col [10000,2]
self.anchor_xmin = [self.temporal_gap * (i-0.5) for i in range(self.temporal_scale)] # [-0.5/100,0.5/100,...98.5/100]
self.anchor_xmax = [self.temporal_gap * (i+0.5) for i in range(1, self.temporal_scale + 1)] # [1.5/100,...,100.5/100]
def _load_file(self, index):
video_name = self.video_list[index]
video_df = pd.read_csv(self.feature_path + "csv_mean_" + str(self.temporal_scale) + "/" + video_name + ".csv")
video_data = video_df.values[:, :]
video_data = torch.Tensor(video_data)
video_data = torch.transpose(video_data, 0, 1)
video_data.float()
return video_data
def _get_train_label(self, index, anchor_xmin, anchor_xmax):
video_name = self.video_list[index] # video_name
video_info = self.video_dict[video_name]
video_frame = video_info['duration_frame']
video_second = video_info['duration_second']
feature_frame = video_info['feature_frame']
corrected_second = float(feature_frame) / video_frame * video_second # there are some frames not used
video_labels = video_info['annotations'] # the measurement is second, not frame
##############################################################################################
# change the measurement from second to percentage
gt_bbox = []
gt_iou_map = []
for j in range(len(video_labels)):
tmp_info = video_labels[j]
tmp_start = max(min(1, tmp_info['segment'][0] / corrected_second), 0)
tmp_end = max(min(1, tmp_info['segment'][1] / corrected_second), 0)
gt_bbox.append([tmp_start, tmp_end]) # gt_bbox [0~1]
tmp_gt_iou_map = iou_with_anchors(
self.match_map[:, 0], self.match_map[:, 1], tmp_start, tmp_end) # [100*100]
tmp_gt_iou_map = np.reshape(tmp_gt_iou_map,
[self.temporal_scale, self.temporal_scale])
gt_iou_map.append(tmp_gt_iou_map)
gt_iou_map = np.array(gt_iou_map) # gt [100*100]
gt_iou_map = np.max(gt_iou_map, axis=0)
gt_iou_map = torch.Tensor(gt_iou_map) # [100,100]
##############################################################################################
####################################################################################################
# generate R_s and R_e
gt_bbox = np.array(gt_bbox) # gt [start,end]
gt_xmins = gt_bbox[:, 0]
gt_xmaxs = gt_bbox[:, 1]
gt_lens = gt_xmaxs - gt_xmins
gt_len_small = 3 * self.temporal_gap # np.maximum(self.temporal_gap, self.boundary_ratio * gt_lens)
gt_start_bboxs = np.stack((gt_xmins - gt_len_small / 2, gt_xmins + gt_len_small / 2), axis=1)
gt_end_bboxs = np.stack((gt_xmaxs - gt_len_small / 2, gt_xmaxs + gt_len_small / 2), axis=1)
#####################################################################################################
##########################################################################################################
# calculate the ioa for all timestamp
match_score_start = []
for jdx in range(len(anchor_xmin)):
match_score_start.append(np.max(
ioa_with_anchors(anchor_xmin[jdx], anchor_xmax[jdx], gt_start_bboxs[:, 0], gt_start_bboxs[:, 1])))
match_score_end = []
for jdx in range(len(anchor_xmin)):
match_score_end.append(np.max(
ioa_with_anchors(anchor_xmin[jdx], anchor_xmax[jdx], gt_end_bboxs[:, 0], gt_end_bboxs[:, 1])))
match_score_start = torch.Tensor(match_score_start)
match_score_end = torch.Tensor(match_score_end)
############################################################################################################
return match_score_start, match_score_end, gt_iou_map
def __len__(self):
return len(self.video_list)
class VideoDataSet_unlabel(data.Dataset):
def __init__(self, opt, subset="unlabel"):
self.temporal_scale = opt["temporal_scale"] # 100
self.temporal_gap = 1. / self.temporal_scale
self.subset = subset
self.mode = opt["mode"]
self.feature_path = opt["feature_path"]
self.video_info_path = "./data/activitynet_annotations/video_info_new_{}.csv".format(opt['unlabel_percent'])
self.video_anno_path = opt["video_anno"]
self._getDatasetDict()
self.unlabel_percent = opt['unlabel_percent']
self._get_match_map()
def _getDatasetDict(self):
anno_df = pd.read_csv(self.video_info_path)
anno_database = load_json(self.video_anno_path)
self.video_dict = {}
for i in range(len(anno_df)):
video_name = anno_df.video.values[i]
video_info = anno_database[video_name]
video_subset = anno_df.subset.values[i]
if self.subset in video_subset:
self.video_dict[video_name] = 'unseen'
self.video_list = list(self.video_dict.keys())
print("%s unlabeled subset video numbers: %d" % (self.subset, len(self.video_list)))
def __getitem__(self, index):
video_data = self._load_file(index)
if self.mode == "train":
# match_score_start, match_score_end, confidence_score = self._get_train_label(index, self.anchor_xmin,
# self.anchor_xmax)
return video_data # ,confidence_score, match_score_start, match_score_end # [400,100],[100,100],[100]
else:
return index, video_data
def _get_match_map(self):
match_map = []
for idx in range(self.temporal_scale):
tmp_match_window = []
xmin = self.temporal_gap * idx
for jdx in range(1, self.temporal_scale + 1):
xmax = xmin + self.temporal_gap * jdx
tmp_match_window.append([xmin, xmax])
match_map.append(tmp_match_window)
match_map = np.array(match_map) # 100x100x2
match_map = np.transpose(match_map, [1, 0, 2]) # [0,1] [1,2] [2,3].....[99,100]
match_map = np.reshape(match_map, [-1, 2]) # [0,2] [1,3] [2,4].....[99,101] # duration x start
self.match_map = match_map # duration is same in row, start is same in col [10000,2]
self.anchor_xmin = [self.temporal_gap * (i-0.5) for i in range(self.temporal_scale)] # [-0.5/100,0.5/100,...98.5/100]
self.anchor_xmax = [self.temporal_gap * (i+0.5) for i in range(1, self.temporal_scale + 1)] # [1.5/100,...,100.5/100]
def _load_file(self, index):
video_name = self.video_list[index]
video_df = pd.read_csv(self.feature_path + "csv_mean_" + str(self.temporal_scale) + "/" + video_name + ".csv")
video_data = video_df.values[:, :]
video_data = torch.Tensor(video_data)
video_data = torch.transpose(video_data, 0, 1)
video_data.float()
return video_data
def _get_train_label(self, index, anchor_xmin, anchor_xmax):
video_name = self.video_list[index] # video_name
video_info = self.video_dict[video_name]
video_frame = video_info['duration_frame']
video_second = video_info['duration_second']
feature_frame = video_info['feature_frame']
corrected_second = float(feature_frame) / video_frame * video_second # there are some frames not used
video_labels = video_info['annotations'] # the measurement is second, not frame
##############################################################################################
# change the measurement from second to percentage
gt_bbox = []
gt_iou_map = []
for j in range(len(video_labels)):
tmp_info = video_labels[j]
tmp_start = max(min(1, tmp_info['segment'][0] / corrected_second), 0)
tmp_end = max(min(1, tmp_info['segment'][1] / corrected_second), 0)
gt_bbox.append([tmp_start, tmp_end]) # gt_bbox [0~1]
tmp_gt_iou_map = iou_with_anchors(
self.match_map[:, 0], self.match_map[:, 1], tmp_start, tmp_end) # [100*100]
tmp_gt_iou_map = np.reshape(tmp_gt_iou_map,
[self.temporal_scale, self.temporal_scale])
gt_iou_map.append(tmp_gt_iou_map)
gt_iou_map = np.array(gt_iou_map) # gt个[100*100]
gt_iou_map = np.max(gt_iou_map, axis=0)
gt_iou_map = torch.Tensor(gt_iou_map) # [100,100]
##############################################################################################
####################################################################################################
# generate R_s and R_e
gt_bbox = np.array(gt_bbox) # gt个[start,end]
gt_xmins = gt_bbox[:, 0]
gt_xmaxs = gt_bbox[:, 1]
gt_lens = gt_xmaxs - gt_xmins
gt_len_small = 3 * self.temporal_gap # np.maximum(self.temporal_gap, self.boundary_ratio * gt_lens)
gt_start_bboxs = np.stack((gt_xmins - gt_len_small / 2, gt_xmins + gt_len_small / 2), axis=1)
gt_end_bboxs = np.stack((gt_xmaxs - gt_len_small / 2, gt_xmaxs + gt_len_small / 2), axis=1)
#####################################################################################################
##########################################################################################################
# calculate the ioa for all timestamp
match_score_start = []
for jdx in range(len(anchor_xmin)):
match_score_start.append(np.max(
ioa_with_anchors(anchor_xmin[jdx], anchor_xmax[jdx], gt_start_bboxs[:, 0], gt_start_bboxs[:, 1])))
match_score_end = []
for jdx in range(len(anchor_xmin)):
match_score_end.append(np.max(
ioa_with_anchors(anchor_xmin[jdx], anchor_xmax[jdx], gt_end_bboxs[:, 0], gt_end_bboxs[:, 1])))
match_score_start = torch.Tensor(match_score_start)
match_score_end = torch.Tensor(match_score_end)
############################################################################################################
return match_score_start, match_score_end, gt_iou_map
def __len__(self):
return len(self.video_list)
if __name__ == '__main__':
import opts
opt = opts.parse_opt()
opt = vars(opt)
train_loader = torch.utils.data.DataLoader(VideoDataSet(opt, subset="train"),
batch_size=opt["batch_size"], shuffle=True,
num_workers=8, pin_memory=True)
for aaa,bbb,ccc,ddd in train_loader: # len(train_loader)=604
set_trace()
print(aaa.shape,bbb.shape,ccc.shape,ddd.shape) # torch.Size([16, 400, 100]) torch.Size([16, 100, 100]) torch.Size([16, 100]) torch.Size([16, 100])
# set_trace()
break