Human-Falling-Detect-Tracks
Human-Falling-Detect-Tracks copied to clipboard
GajuuzZ
Original dataset annotation 原始数据集标注
I used the Le2i Fall detection dataset downloaded from the Internet. Is the label of this dataset (the behavior category of each frame of video) manually marked or how to handle it 我使用了网上下载的跌倒数据集Le2i Fall detection Dataset ，对于这个数据集的标签（每一帧视频的行为类别）是自己手动标注吗，还是怎么处理
in this file video(1).txt ， What does the second column represent? Is it the behavior category of people in each frame of video
在此中 video(1).txt文件中，第二列代表什么？这是视频每一帧中人的行为类别吗
48
80
1,1,0,0,0,0
2,1,0,0,0,0
3,1,0,0,0,0
4,1,0,0,0,0
5,1,292,152,311,240
6,1,292,152,311,240
7,1,292,152,311,240
8,1,285,131,315,241
9,1,284,114,315,239
10,1,284,114,315,239
11,1,279,68,312,241
12,1,275,63,315,240
13,1,268,66,316,241
14,1,268,66,316,241
15,1,258,64,314,241
16,1,244,61,315,240
17,1,234,66,313,239
18,1,228,59,312,240
在 create_dataset2.py中 传入 '../Data/Coffee_room_new-pose+score.csv' 文件中的 label 值是从下面的代码取出的 ` # Pose Labels. frames_label = annot[annot['video'] == vid].reset_index(drop=True)
cap = cv2.VideoCapture(os.path.join(video_folder, vid))
frames_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
frame_size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
Bounding Boxs Labels.
annot_file = os.path.join(annot_folder, vid.split('.')[0], '.txt')
annot = None
if os.path.exists(annot_file):
annot = pd.read_csv(annot_file, header=None,
names=['frame_idx', 'class', 'xmin', 'ymin', 'xmax', 'ymax'])
annot = annot.dropna().reset_index(drop=True)
assert frames_count == len(annot), 'frame count not equal! {} and {}'.format(frames_count, len(annot))
fps_time = 0
i = 1
while True:
ret, frame = cap.read()
if ret:
frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
cls_idx = int(frames_label[frames_label['frame'] == > i]['label'])`
但是 从 # Pose Labels中提取的 label 值都是 一样的，都是 cls_idx=0，在 create_dataset3中又如何进行下一步的操作呢， 还有就是 # Bounding Boxs Labels.中的 “ class”代表着什么呢.
