划分数据集

数据集存在于0none 　1pouting 　2smile　3openmouth四个文件夹中，根据要求，我们对其按照 1:9 的比例进行划分
注意，这里我们对每一个文件夹下的数据按比例进行划分，保证数据的均匀

import os
import random
import shutil
from shutil import copy2

foldernames = ['0none', '1pouting', '2smile', '3openmouth']
trainDir = 'train/'
valDir = 'val/'

for foldername in foldernames:
    datafile = os.listdir(foldername) 
    datafilename = foldername + '/'
    num_train = len(datafile) 
    print("Images in " + foldername + " : " + str(num_train))
    index_list = list(range(num_train))
#    print(index_list)d
    num = 0

    for i in index_list:
    filename = os.path.join(datafilename, datafile[i])    
    if num < num_train * 0.9:
        copy2(filename, trainDir)
    else:
        copy2(filename, valDir)
    
    num += 1

分割完成后，得到train val两个文件夹，分别存储训练和测试的图片数据

image.png

查看训练集的图片数

image.png

查看测试集的图片数

image.png
对数据集进行划分后，还有对数据设置标签

#!/usr/bin/env sh

DATA=dataset
DATA_TRAIN=dataset/train
DATA_VAL=dataset/val

echo "Creating train.txt ..."
rm -rf $DATA/train.txt
find $DATA_TRAIN -name *none.jpg | cut -d '/' -f3 | sed "s/$/ 0/">>$DATA/train.txt

find $DATA_TRAIN -name *pouting.jpg | cut -d '/' -f3 | sed "s/$/ 1/">>$DATA/train.txt

find $DATA_TRAIN -name *smile.jpg | cut -d '/' -f3 | sed "s/$/ 2/">>$DATA/train.txt

find $DATA_TRAIN -name *open.jpg | cut -d '/' -f3 | sed "s/$/ 3/">>$DATA/train.txt
echo "Creat train.txt is done"

echo "Creating test.txt ..."
rm -rf $DATA/val.txt
find $DATA_VAL -name *none.jpg | cut -d '/' -f3 | sed "s/$/ 0/">>$DATA/val.txt

find $DATA_VAL -name *pouting.jpg | cut -d '/' -f3 | sed "s/$/ 1/">>$DATA/val.txt

find $DATA_VAL -name *smile.jpg | cut -d '/' -f3 | sed "s/$/ 2/">>$DATA/val.txt

find $DATA_VAL -name *open.jpg | cut -d '/' -f3 | sed "s/$/ 3/">>$DATA/val.txt
echo "Creat val.txt is done"

image.png