在使用百度飞桨下PaddleOCR2.6训练识别部分时，官方教程在项目下./doc/doc_ch/recognition.md,详细介绍了识别训练，具体过程可以参照此md文件

说一下遇到的问题，ICDAR2015数据格式如下：

 两个out.txt文件是输出的，原有的.txt文件虽然集合了图片路径和标注信息，但给的路径是相对路径，在训练识别部分时会报错，imgs does not exist  很让人崩溃是不是。

 在官方转换数据格式代码的基础上做出修改，路径在./ppocr/utils/gen_label.py,官方代码：

import os
import argparse
import json


def gen_rec_label(input_path, out_label):
    with open(out_label, 'w') as out_file:
        with open(input_path, 'r') as f:
            for line in f.readlines():
                tmp = line.strip('\n').replace(" ", "").split(',')
                img_path, label = tmp[0], tmp[1]
                label = label.replace("\"", "")
                out_file.write(img_path + '\t' + label + '\n')


def gen_det_label(root_path, input_dir, out_label):
    with open(out_label, 'w') as out_file:
        for label_file in os.listdir(input_dir):
            img_path = root_path + label_file[3:-4] + ".jpg"
            label = []
            with open(
                    os.path.join(input_dir, label_file), 'r',
                    encoding='utf-8-sig') as f:
                for line in f.readlines():
                    tmp = line.strip("\n\r").replace("\xef\xbb\xbf",
                                                     "").split(',')
                    points = tmp[:8]
                    s = []
                    for i in range(0, len(points), 2):
                        b = points[i:i + 2]
                        b = [int(t) for t in b]
                        s.append(b)
                    result = {"transcription": tmp[8], "points": s}
                    label.append(result)

            out_file.write(img_path + '\t' + json.dumps(
                label, ensure_ascii=False) + '\n')


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument(
        '--mode',
        type=str,
        default="rec",
        help='Generate rec_label or det_label, can be set rec or det')
    parser.add_argument(
        '--root_path',
        type=str,
        default=".",
        help='The root directory of images.Only takes effect when mode=det ')
    parser.add_argument(
        '--input_path',
        type=str,
        default="D:/Learn/paddleocr2.6/train_data/icdar2015/text_localization/test_icdar2015_label.txt",
        help='Input_label or input path to be converted')
    parser.add_argument(
        '--output_label',
        type=str,
        default="out_rec_test_label.txt",
        help='Output file name')

    args = parser.parse_args()
    if args.mode == "rec":
        print("Generate rec label")
        gen_rec_label(args.input_path, args.output_label)
    elif args.mode == "det":
        gen_det_label(args.root_path, args.input_path, args.output_label)

修改之后的代码：

import os
import argparse
import json


def collect_paths_and_annotations(input_path, output_label, mode):
    with open(output_label, 'w') as out_file:
        with open(input_path, 'r') as f:
            for line in f.readlines():
                tmp = line.strip().split('\t')
                img_path, annotations = tmp[0], tmp[1]
                img_path = os.path.abspath(img_path)  # Convert to absolute path
                img_path = img_path.replace("\\", "/")  # Convert to Unix-style path

                if mode == "rec":
                    # In rec mode, annotations are the labels directly
                    out_file.write(f"{img_path}\t{annotations}\n")
                elif mode == "det":
                    # In det mode, annotations are in JSON format, so we parse them
                    annotations = json.loads(annotations)
                    annotations_str = json.dumps(annotations, ensure_ascii=False)
                    out_file.write(f"{img_path}\t{annotations_str}\n")


if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument(
        '--mode',
        type=str,
        default="rec",
        help='Generate rec_label or det_label, can be set rec or det')
    parser.add_argument(
        '--input_path',
        type=str,
        default="D:/Learn/paddleocr2.6/train_data/icdar2015/text_localization/train_icdar2015_label.txt",
        help='Input_label or input path to be converted')
    parser.add_argument(
        '--output_label',
        type=str,
        default="D:/Learn/paddleocr2.6/train_data/out_rec_train_label.txt",
        help='Output file name')

    args = parser.parse_args()
    collect_paths_and_annotations(args.input_path, args.output_label, args.mode)

这样把原有的相对路径.txt文件转换成绝对路径

问题解决！

顺便一提opencv-python版本兼容的问题，最新版本opencv会和项目有一定冲突，经过实验，发现以下版本最好 

pip install opencv-python==4.6.0.66

 顺利训练！