数据清洗操作步骤
1、txt第一行删掉
2、txt中null替换为0,0,0,0,0,0（之后会删除这些帧）
3、运行data_fixation.py
#txt文件名改为数字,删第一行，将null替换为0,0,0,0,0,0,输入此程序，可获得每帧fixation点的重心
import os
import numpy as np
from numpy import *
path = "./test/"#输入文件的绝对路径
lists = os.listdir(path) #获取目录下的所有文件名
for list in lists: #遍历文件名
txt_dir = os.path.join(path,list) #每个txt文件的路径
position = path+list
print(position)
with open(position, 'r+') as f:
for lines in f.readlines():
#print(lines.strip())
items = lines.strip().split(',')
retName = [] # 列表 第一列，每帧图像名字
retName.append(items[0]) # 第一列，每帧图像名字
x = [] # 第一个fixation,列表
y = [] # 最后fixation,列表
m = len(items)
print(m)
x.append([float(items[i]) for i in range(6, m-1, 2)])
y.append([float(items[i]) for i in range(7, m, 2)])
x1 = "".join([str(i)for i in x]) # join转为字符串
x2=x1.replace('[','') #去除方括号
x3=x2.replace(']','')
y1 = "".join([str(i)for i in y]) # join转为字符串
y2=y1.replace('[','') #去除方括号
y3=y2.replace(']','')
xx = x3.strip().split(',')
yy = y3.strip().split(',')
x.append([float(xx[i]) for i in range(0,len(xx))])
y.append([float(yy[i]) for i in range(0, len(yy))])
a = sum(x)/2
b = sum(y)/2
lx = len(xx)
ly = len(yy)
nx = a / lx #fixation重心x值
ny = b / ly #fixation重心y值
path1 = "C:/data/hero_fixation/482/" # 生成每个txt的绝对路径
retName1 = "".join(retName) #join转为字符串
#print(retName1)
position1 = path1 + retName1 + '.txt' # 新建每个txt文件的路径
#print(position1)
with open(position1, 'a') as f1:
f1.write(str(nx)+",")
f1.write(str(ny))获得以第一列字符串命名的txt文件，txt中存放此帧fixation点的重心。其中，test中为输入txt,data中为输出txt集。
4、运行data_action.py
#txt文件名改为数字,删第一行，将null替换为0,0,0,0,0,0,输入此程序，可获得每帧action
import os
path = "./test/"#输入文件的绝对路径
lists = os.listdir(path) #获取目录下的所有文件名
for list in lists: #遍历文件名
txt_dir = os.path.join(path,list) #每个txt文件的路径
position = path+list
print(position)
with open(position, 'r+') as f:
for lines in f.readlines():
#print(lines.strip())
items = lines.strip().split(',')
retName = [] # 列表 第一列，每帧图像名字
retName.append(items[0]) # 第一列，每帧图像名字
action = []
action.append([int(items[i]) for i in range(5,6)])
action1 = "".join([str(i)for i in action]) # join转为字符串
action2=action1.replace('[','') #去除方括号
action3=action2.replace(']','')
path1 = "F:/ROII/cnn_yolov5/data/hero_action/train/482_RZ_3472304_Jul-18-12-52-35/" # 生成每个txt的绝对路径
retName1 = "".join(retName) #join转为字符串
#print(retName1)
position1 = path1 + retName1 + '.txt' # 新建每个txt文件的路径
#print(position1)
with open(position1, 'a') as f1:
f1.write(action3+'\n')获得以第一列字符串命名的txt文件，txt中存放此帧action值。其中，test中为输入txt,data中为输出txt集。
5、删除不存在fixation及fixation超出边界的帧
复制2之后txt,
运行data_delete_null.py
#data_delete_null.py
#txt文件名改为数字,删第一行，将null替换为0,0,0,0,0,0,输入此程序，可获得重心为0的帧的名称，再次输入data_null_bat.py可生成删除的bat命令
import os
import numpy as np
from numpy import *
path = "./test/"#输入文件的绝对路径
lists = os.listdir(path) #获取目录下的所有文件名
for list in lists: #遍历文件名
txt_dir = os.path.join(path,list) #每个txt文件的路径
position = path+list
print(position)
print(list)
with open(position, 'r+') as f:
for lines in f.readlines():
#print(lines.strip())
items = lines.strip().split(',')
retName = [] # 列表 第一列，每帧图像名字
retName.append(items[0]) # 第一列，每帧图像名字
x = [] # 第一个fixation,列表
y = [] # 最后fixation,列表
m = len(items)
#print(m)
x.append([float(items[i]) for i in range(6, m-1, 2)])
y.append([float(items[i]) for i in range(7, m, 2)])
x1 = "".join([str(i)for i in x]) # join转为字符串
x2=x1.replace('[','') #去除方括号
x3=x2.replace(']','')
y1 = "".join([str(i)for i in y]) # join转为字符串
y2=y1.replace('[','') #去除方括号
y3=y2.replace(']','')
xx = x3.strip().split(',')
yy = y3.strip().split(',')
x.append([float(xx[i]) for i in range(0,len(xx))])
y.append([float(yy[i]) for i in range(0, len(yy))])
a = sum(x)/2
b = sum(y)/2
lx = len(xx)
ly = len(yy)
nx = a / lx #fixation重心x值
ny = b / ly #fixation重心y值
path1 = "./Anullout/" # 生成每个txt的绝对路径
retName1 = "".join(retName) #join转为字符串
#print(retName1)
#position1 = path1 + retName1 + '.txt' # 新建每个txt文件的路径
position1 = path1 + list # 新建每个txt文件的路径
if nx==0:
with open(position1, 'a') as f1:
f1.write(str(retName1) + "\n")
elif ny==0:
with open(position1, 'a') as f1:
f1.write(str(retName1) + "\n")
elif nx>160:
with open(position1, 'a') as f1:
f1.write(str(retName1) + "\n")
elif ny>210:
with open(position1, 'a') as f1:
f1.write(str(retName1) + "\n")生成剔除帧的编号，改为数字后，输入data_null_bat.py改为bat命令
#data_null_bat.py
#将data_deleta_null.py生成输入data_null_bat.py可生成删除的bat命令
import os
import re
import sys
from numpy import *
path = "./Anullout/"#输入文件的绝对路径
lists = os.listdir(path) #获取目录下的所有文件名
for list in lists: #遍历文件名
txt_dir = os.path.join(path,list) #每个txt文件的路径
position = path+list
print(position)
print(list)
with open(position, 'r+') as f:
for lines in f.readlines():
#print(lines.strip())
line = lines[:-1]
#print(line.zfill(7))
path1 = "./Anullout/bat/" # 输出文件的绝对路径
position1 = path1 + list # 新建每个txt文件的路径
rs=line.zfill(7) #每个序号改为0000001,0000002
newname = rs.replace(rs, 'del' + ' ' + rs + '.png')
with open(position1, 'a') as f1:
f1.write(str(newname + "\n"))