gen_ocr_train_val_test.py 6.0 KB
Newer Older
1 2 3 4 5 6
# coding:utf8
import os
import shutil
import random
import argparse

7

–
MrCuiHao 已提交
8
# 删除划分的训练集、验证集、测试集文件夹,重新创建一个空的文件夹
9 10
def isCreateOrDeleteFolder(path, flag):
    flagPath = os.path.join(path, flag)
–
MrCuiHao 已提交
11

12 13
    if os.path.exists(flagPath):
        shutil.rmtree(flagPath)
–
MrCuiHao 已提交
14

15 16 17 18 19
    os.makedirs(flagPath)
    flagAbsPath = os.path.abspath(flagPath)
    return flagAbsPath


–
MrCuiHao 已提交
20 21
def splitTrainVal(root, dir, absTrainRootPath, absValRootPath, absTestRootPath, trainTxt, valTxt, testTxt, flag):
    # 按照指定的比例划分训练集、验证集、测试集
22 23
    labelPath = os.path.join(root, dir)
    labelAbsPath = os.path.abspath(labelPath)
–
MrCuiHao 已提交
24

25 26 27 28
    if flag == "det":
        labelFilePath = os.path.join(labelAbsPath, args.detLabelFileName)
    elif flag == "rec":
        labelFilePath = os.path.join(labelAbsPath, args.recLabelFileName)
–
MrCuiHao 已提交
29

30 31 32 33
    labelFileRead = open(labelFilePath, "r", encoding="UTF-8")
    labelFileContent = labelFileRead.readlines()
    random.shuffle(labelFileContent)
    labelRecordLen = len(labelFileContent)
–
MrCuiHao 已提交
34

35 36 37 38
    for index, labelRecordInfo in enumerate(labelFileContent):
        imageRelativePath = labelRecordInfo.split('\t')[0]
        imageLabel = labelRecordInfo.split('\t')[1]
        imageName = os.path.basename(imageRelativePath)
–
MrCuiHao 已提交
39

40 41 42 43
        if flag == "det":
            imagePath = os.path.join(labelAbsPath, imageName)
        elif flag == "rec":
            imagePath = os.path.join(labelAbsPath, "{}\\{}".format(args.recImageDirName, imageName))
–
MrCuiHao 已提交
44 45 46 47 48 49 50 51

        # 按预设的比例划分训练集、验证集、测试集
        trainValTestRatio = args.trainValTestRatio.split(":")
        trainRatio = eval(trainValTestRatio[0]) / 10
        valRatio = trainRatio + eval(trainValTestRatio[1]) / 10
        curRatio = index / labelRecordLen

        if curRatio < trainRatio:
52 53 54
            imageCopyPath = os.path.join(absTrainRootPath, imageName)
            shutil.copy(imagePath, imageCopyPath)
            trainTxt.write("{}\t{}".format(imageCopyPath, imageLabel))
–
MrCuiHao 已提交
55
        elif curRatio >= trainRatio and curRatio < valRatio:
56 57 58
            imageCopyPath = os.path.join(absValRootPath, imageName)
            shutil.copy(imagePath, imageCopyPath)
            valTxt.write("{}\t{}".format(imageCopyPath, imageLabel))
–
MrCuiHao 已提交
59 60 61 62
        else:
            imageCopyPath = os.path.join(absTestRootPath, imageName)
            shutil.copy(imagePath, imageCopyPath)
            testTxt.write("{}\t{}".format(imageCopyPath, imageLabel))
63 64


65 66 67 68 69 70
# 删掉存在的文件
def removeFile(path):
    if os.path.exists(path):
        os.remove(path)


71 72 73
def genDetRecTrainVal(args):
    detAbsTrainRootPath = isCreateOrDeleteFolder(args.detRootPath, "train")
    detAbsValRootPath = isCreateOrDeleteFolder(args.detRootPath, "val")
–
MrCuiHao 已提交
74
    detAbsTestRootPath = isCreateOrDeleteFolder(args.detRootPath, "test")
75 76
    recAbsTrainRootPath = isCreateOrDeleteFolder(args.recRootPath, "train")
    recAbsValRootPath = isCreateOrDeleteFolder(args.recRootPath, "val")
–
MrCuiHao 已提交
77 78
    recAbsTestRootPath = isCreateOrDeleteFolder(args.recRootPath, "test")

79 80
    removeFile(os.path.join(args.detRootPath, "train.txt"))
    removeFile(os.path.join(args.detRootPath, "val.txt"))
–
MrCuiHao 已提交
81
    removeFile(os.path.join(args.detRootPath, "test.txt"))
82 83
    removeFile(os.path.join(args.recRootPath, "train.txt"))
    removeFile(os.path.join(args.recRootPath, "val.txt"))
–
MrCuiHao 已提交
84 85
    removeFile(os.path.join(args.recRootPath, "test.txt"))

86 87
    detTrainTxt = open(os.path.join(args.detRootPath, "train.txt"), "a", encoding="UTF-8")
    detValTxt = open(os.path.join(args.detRootPath, "val.txt"), "a", encoding="UTF-8")
–
MrCuiHao 已提交
88
    detTestTxt = open(os.path.join(args.detRootPath, "test.txt"), "a", encoding="UTF-8")
89 90
    recTrainTxt = open(os.path.join(args.recRootPath, "train.txt"), "a", encoding="UTF-8")
    recValTxt = open(os.path.join(args.recRootPath, "val.txt"), "a", encoding="UTF-8")
–
MrCuiHao 已提交
91 92
    recTestTxt = open(os.path.join(args.recRootPath, "test.txt"), "a", encoding="UTF-8")

93 94
    for root, dirs, files in os.walk(args.labelRootPath):
        for dir in dirs:
–
MrCuiHao 已提交
95 96 97 98
            splitTrainVal(root, dir, detAbsTrainRootPath, detAbsValRootPath, detAbsTestRootPath, detTrainTxt, detValTxt,
                          detTestTxt, "det")
            splitTrainVal(root, dir, recAbsTrainRootPath, recAbsValRootPath, recAbsTestRootPath, recTrainTxt, recValTxt,
                          recTestTxt, "rec")
99 100 101 102
        break


if __name__ == "__main__":
–
MrCuiHao 已提交
103
    # 功能描述:分别划分检测和识别的训练集、验证集、测试集
104
    # 说明:可以根据自己的路径和需求调整参数,图像数据往往多人合作分批标注,每一批图像数据放在一个文件夹内用PPOCRLabel进行标注,
–
MrCuiHao 已提交
105
    # 如此会有多个标注好的图像文件夹汇总并划分训练集、验证集、测试集的需求
106 107
    parser = argparse.ArgumentParser()
    parser.add_argument(
–
MrCuiHao 已提交
108 109 110 111
        "--trainValTestRatio",
        type=str,
        default="6:2:2",
        help="ratio of trainset:valset:testset")
112 113 114
    parser.add_argument(
        "--labelRootPath",
        type=str,
–
MrCuiHao 已提交
115
        default="../train_data/label",
116 117 118 119 120
        help="path to the dataset marked by ppocrlabel, E.g, dataset folder named 1,2,3..."
    )
    parser.add_argument(
        "--detRootPath",
        type=str,
–
MrCuiHao 已提交
121
        default="../train_data/det",
122 123 124 125
        help="the path where the divided detection dataset is placed")
    parser.add_argument(
        "--recRootPath",
        type=str,
–
MrCuiHao 已提交
126
        default="../train_data/rec",
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
        help="the path where the divided recognition dataset is placed"
    )
    parser.add_argument(
        "--detLabelFileName",
        type=str,
        default="Label.txt",
        help="the name of the detection annotation file")
    parser.add_argument(
        "--recLabelFileName",
        type=str,
        default="rec_gt.txt",
        help="the name of the recognition annotation file"
    )
    parser.add_argument(
        "--recImageDirName",
        type=str,
        default="crop_img",
        help="the name of the folder where the cropped recognition dataset is located"
    )
    args = parser.parse_args()
    genDetRecTrainVal(args)