End-to-End训练自定义数据集总是报错

[Jerry200404]

目的：1.使用 JSON 格式标注，并在 type 字段区分 手写体 和 印刷体，以达到分类效果。
2.训练检测模型，使用 polygon 标注 单词位置，然后传入识别模型，所以采用End-to-End。

俄语数据集文件目录结构
PP-OCRv4-dataset/
├── train/
│ ├── annotations/
│ │ ├── 0_0.json
│ │ ├── 0_1.json
│ │ ├── 0_2.json
│ ├── images/
│ │ ├── 0_0/
│ │ │ ├── img1.jpg
│ │ │ ├── img2.jpg
│ │ ├── 0_1/
│ │ │ ├── img3.jpg
│ │ │ ├── img4.jpg
test同理未列出
json文件内部数据格式：
{
"image": "images/0_0/lhr_0_1000_01184616.jpg",
"annotations": [
{
"polygon": [0.58, 0.72, 0.86, 0.72, 0.86, 0.96, 0.58, 0.96],
"label": "воссоздать",
"type": "printed"
},
{
"polygon": [0.12, 0.35, 0.46, 0.35, 0.46, 0.62, 0.12, 0.62],
"label": "коммунизм",
"type": "handwriting"
}
]
}

修改了e2e_r50_vd_pg.yml配置文件：
`Global:
use_gpu: True
epoch_num: 600
log_smooth_window: 20
print_batch_step: 10
save_model_dir: ./output/pgnet_r50_vd_totaltext/
save_epoch_step: 10

evaluation is run every 0 iterationss after the 1000th iteration

eval_batch_step: [ 0, 1000 ]
cal_metric_during_train: False
pretrained_model:
checkpoints:
save_inference_dir:
use_visualdl: False
infer_img:
infer_visual_type: RU # two mode: EN is for english datasets, CN is for chinese datasets
valid_set: totaltext # two mode: totaltext valid curved words, partvgg valid non-curved words
save_res_path: ./output/pgnet_r50_vd_totaltext/predicts_pgnet.txt
character_dict_path: ppocr/utils/dict/cyrillic_dict.txt
character_type: RU
max_text_length: 50 # the max length in seq
max_text_nums: 30 # the max seq nums in a pic
tcl_len: 64

Architecture:
model_type: e2e
algorithm: PGNet
Transform:
Backbone:
name: ResNet
layers: 50
Neck:
name: PGFPN
Head:
name: PGHead
character_dict_path: ppocr/utils/dict/cyrillic_dict.txt # the same as Global:character_dict_path

Loss:
name: PGLoss
tcl_bs: 64
max_text_length: 50 # the same as Global: max_text_length
max_text_nums: 30 # the same as Global：max_text_nums
pad_num: 36 # the length of dict for pad

Optimizer:
name: Adam
beta1: 0.9
beta2: 0.999
lr:
name: Cosine
learning_rate: 0.001
warmup_epoch: 50
regularizer:
name: 'L2'
factor: 0.0001

PostProcess:
name: PGPostProcess
score_thresh: 0.5
mode: fast # fast or slow two ways
point_gather_mode: align # same as PGProcessTrain: point_gather_mode

Metric:
name: E2EMetric
mode: A # two ways for eval, A: label from txt, B: label from gt_mat
gt_mat_dir: ./train_data/total_text/gt # the dir of gt_mat
character_dict_path: ppocr/utils/dict/cyrillic_dict.txt
main_indicator: f_score_e2e

Train:
dataset:
name: SimpleDataSet #PGDataSet
data_dir: ./PP-OCRv4-dataset/train/
label_file_list: ["annotations/*.json"]
ratio_list: [1.0]
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- E2ELabelEncodeTrain:
- PGProcessTrain:
batch_size: 14 # same as loader: batch_size_per_card
use_resize: True
use_random_crop: False
min_crop_size: 24
min_text_size: 4
max_text_size: 512
point_gather_mode: align # two mode: align and none, align mode is better than none mode
- KeepKeys:
keep_keys: [ 'images', 'tcl_maps', 'tcl_label_maps', 'border_maps','direction_maps', 'training_masks', 'label_list', 'pos_list', 'pos_mask' ] # dataloader will return list in this order
loader:
shuffle: True
drop_last: True
batch_size_per_card: 14
num_workers: 16

Eval:
dataset:
name: SimpleDataSet
data_dir: ./PP-OCRv4-dataset/test/
label_file_list: ["annotations/*.json"]
transforms:
- DecodeImage: # load image
img_mode: BGR
channel_first: False
- E2ELabelEncodeTest:
- E2EResizeForTest:
max_side_len: 768
- NormalizeImage:
scale: 1./255.
mean: [ 0.485, 0.456, 0.406 ]
std: [ 0.229, 0.224, 0.225 ]
order: 'hwc'
- ToCHWImage:
- KeepKeys:
keep_keys: [ 'image', 'shape', 'polys', 'texts', 'ignore_tags', 'img_id']
loader:
shuffle: False
drop_last: False
batch_size_per_card: 1 # must be 1
num_workers: 2
修改了SimpleDataSet用来加载数据：class SimpleDataSet(Dataset):
def init(self, config, mode, logger, seed=None):
super(SimpleDataSet, self).init()
self.logger = logger
self.mode = mode.lower()

    # 获取配置
    global_config = config["Global"]
    dataset_config = config[mode]["dataset"]
    loader_config = config[mode]["loader"]

    # 配置初始化
    self.delimiter = dataset_config.get("delimiter", "\t")
    label_file_list = dataset_config.pop("label_file_list")
    self.data_dir = dataset_config["data_dir"]
    self.do_shuffle = loader_config["shuffle"]
    self.seed = seed

    # 获取数据分割比例
    ratio_list = dataset_config.get("ratio_list", 1.0)
    if isinstance(ratio_list, (float, int)):
        ratio_list = [float(ratio_list)] * len(label_file_list)
    assert len(ratio_list) == len(label_file_list), \
        f"Mismatch: len(ratio_list)={len(ratio_list)}, len(label_file_list)={len(label_file_list)}"

    self.logger.info(f"Initialize indexs of datasets: {label_file_list}")
    self.data_lines = self.get_image_info_list(label_file_list, ratio_list)
    self.data_idx_order_list = list(range(len(self.data_lines)))

    if self.mode == "train" and self.do_shuffle:
        self.shuffle_data_random()

    self.ops = create_operators(dataset_config["transforms"], global_config)
    self.ext_op_transform_idx = dataset_config.get("ext_op_transform_idx", 2)
    self.need_reset = any(x < 1 for x in ratio_list)

def get_image_info_list(self, file_list, ratio_list):
    """ 解析 JSON 标注文件，并构造图片路径和标注信息 """
    data_lines = []
    json_files = glob.glob(os.path.join(self.data_dir, "annotations", "*.json"))

    if not json_files:
        self.logger.error(f"No valid annotation files found in {file_list}")
        return []

    # 调整比例列表
    if len(ratio_list) < len(json_files):
        ratio_list = ratio_list + [ratio_list[-1]] * (len(json_files) - len(ratio_list))
    elif len(ratio_list) > len(json_files):
        ratio_list = ratio_list[:len(json_files)]

    self.logger.info(f"Using ratio list: {ratio_list}")

    for idx, json_path in enumerate(json_files):
        if not os.path.exists(json_path):
            self.logger.error(f"Annotation file not found: {json_path}")
            continue

        # 读取 JSON 文件
        with open(json_path, "r", encoding="windows-1251") as f:
            try:
                annotations = json.load(f)
            except json.JSONDecodeError as e:
                self.logger.error(f"Error parsing JSON file {json_path}: {e}")
                continue

            if not annotations:
                self.logger.warning(f"Empty annotations in {json_path}, skipping.")
                continue

            # 选择样本数据
            ratio_value = ratio_list[idx]
            sample_size = min(round(len(annotations) * ratio_value), len(annotations))
            annotations = random.sample(annotations, sample_size)

            # 处理每个 annotation
            for annotation in annotations:
                image_path = annotation["image"]
                full_image_path = os.path.join(self.data_dir, image_path)

                if not os.path.exists(full_image_path):
                    self.logger.warning(f"Image file not found: {full_image_path}")
                    continue

                # 将 'text' 字段替换为 'label' 并确保没有 text 字段
                processed_annotations = []
                for ann in annotation["annotations"]:
                    # 如果没有 "label" 字段，跳过
                    if "label" not in ann:
                        self.logger.warning(f"Missing 'label' in annotation: {ann}")
                        continue

                    processed_annotations.append(ann)

                data_lines.append({
                    "image": full_image_path,
                    "annotations": processed_annotations
                })

    return data_lines

def shuffle_data_random(self):
    """ 在训练模式下随机打乱数据 """
    random.seed(self.seed)
    random.shuffle(self.data_lines)

def get_ext_data(self):
    """ 额外数据增强（如果需要） """
    ext_data_num = 0
    for op in self.ops:
        if hasattr(op, "ext_data_num"):
            ext_data_num = getattr(op, "ext_data_num")
            break
    load_data_ops = self.ops[:self.ext_op_transform_idx]
    ext_data = []

    while len(ext_data) < ext_data_num:
        file_idx = self.data_idx_order_list[np.random.randint(self.__len__())]
        data_line = self.data_lines[file_idx]
        img_path = data_line["image"]
        data = {"img_path": img_path, "annotations": data_line["annotations"]}

        if not os.path.exists(img_path):
            continue

        with open(img_path, "rb") as f:
            data["image"] = f.read()

        data = transform(data, load_data_ops)
        if data is None:
            continue
        if "polys" in data and data["polys"].shape[1] != 4:
            continue
        ext_data.append(data)

    return ext_data

def __getitem__(self, idx):
    """ 获取数据样本 """
    file_idx = self.data_idx_order_list[idx]
    data_line = self.data_lines[file_idx]

    try:
        img_path = data_line["image"]
        annotations = data_line["annotations"]
        data = {"img_path": img_path, "annotations": annotations}

        if not os.path.exists(img_path):
            raise FileNotFoundError(f"{img_path} does not exist!")

        with open(img_path, "rb") as f:
            data["image"] = f.read()

        processed_annotations = []
        for annotation in annotations:
            polygon = annotation["polygon"]
            label = annotation["label"]  # 获取 label 字段
            annotation_type = annotation["type"]

            processed_annotations.append({
                "polygon": polygon,
                "label": label,
                "type": annotation_type
            })
        data["annotations"] = processed_annotations

        data["ext_data"] = self.get_ext_data()
        data["filename"] = img_path
        outs = transform(data, self.ops)

    except Exception as e:
        self.logger.error(f" Error occurred while parsing {data_line if data_line else 'None'}: {e}")
        outs = None

    if outs is None:
        rnd_idx = np.random.randint(self.__len__()) if self.mode == "train" else (idx + 1) % self.__len__()
        return self.__getitem__(rnd_idx)
    return outs

def __len__(self):
    """ 数据集样本数量 """
    return len(self.data_idx_order_list)`

终端执行命令：
python tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml
训练报错：
训练报错信息.txt
报错部分信息：
PS D:\PP-OCR\PaddleOCR> python tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml
[2025/02/14 20:10:35] ppocr INFO: Architecture :
[2025/02/14 20:10:35] ppocr INFO: Backbone :
[2025/02/14 20:10:35] ppocr INFO: layers : 50
[2025/02/14 20:10:35] ppocr INFO: name : ResNet
[2025/02/14 20:10:35] ppocr INFO: Head :
[2025/02/14 20:10:35] ppocr INFO: character_dict_path : ppocr/utils/dict/cyrillic_dict.txt
[2025/02/14 20:10:35] ppocr INFO: name : PGHead
[2025/02/14 20:10:35] ppocr INFO: Neck :
[2025/02/14 20:10:35] ppocr INFO: name : PGFPN
[2025/02/14 20:10:35] ppocr INFO: Transform : None
[2025/02/14 20:10:35] ppocr INFO: algorithm : PGNet
[2025/02/14 20:10:35] ppocr INFO: model_type : e2e
[2025/02/14 20:10:35] ppocr INFO: Eval :
[2025/02/14 20:10:35] ppocr INFO: dataset :
[2025/02/14 20:10:35] ppocr INFO: data_dir : ./PP-OCRv4-dataset/test/
[2025/02/14 20:10:35] ppocr INFO: label_file_list : ['annotations/.json']
[2025/02/14 20:10:35] ppocr INFO: name : SimpleDataSet
[2025/02/14 20:10:35] ppocr INFO: transforms :
[2025/02/14 20:10:35] ppocr INFO: DecodeImage :
[2025/02/14 20:10:35] ppocr INFO: channel_first : False
[2025/02/14 20:10:35] ppocr INFO: img_mode : BGR
[2025/02/14 20:10:35] ppocr INFO: E2ELabelEncodeTest : None
[2025/02/14 20:10:35] ppocr INFO: E2EResizeForTest :
[2025/02/14 20:10:35] ppocr INFO: max_side_len : 768
[2025/02/14 20:10:35] ppocr INFO: NormalizeImage :
[2025/02/14 20:10:35] ppocr INFO: mean : [0.485, 0.456, 0.406]
[2025/02/14 20:10:35] ppocr INFO: order : hwc
[2025/02/14 20:10:35] ppocr INFO: scale : 1./255.
[2025/02/14 20:10:35] ppocr INFO: std : [0.229, 0.224, 0.225]
[2025/02/14 20:10:35] ppocr INFO: ToCHWImage : None
[2025/02/14 20:10:35] ppocr INFO: KeepKeys :
[2025/02/14 20:10:35] ppocr INFO: keep_keys : ['image', 'shape', 'polys', 'texts', 'ignore_tags', 'img_id']
[2025/02/14 20:10:35] ppocr INFO: loader :
[2025/02/14 20:10:35] ppocr INFO: batch_size_per_card : 1
[2025/02/14 20:10:35] ppocr INFO: drop_last : False
[2025/02/14 20:10:35] ppocr INFO: num_workers : 2
[2025/02/14 20:10:35] ppocr INFO: shuffle : False
[2025/02/14 20:10:35] ppocr INFO: Global :
[2025/02/14 20:10:35] ppocr INFO: cal_metric_during_train : False
[2025/02/14 20:10:35] ppocr INFO: character_dict_path : ppocr/utils/dict/cyrillic_dict.txt
[2025/02/14 20:10:35] ppocr INFO: character_type : RU
[2025/02/14 20:10:35] ppocr INFO: checkpoints : None
[2025/02/14 20:10:35] ppocr INFO: distributed : False
[2025/02/14 20:10:35] ppocr INFO: epoch_num : 600
[2025/02/14 20:10:35] ppocr INFO: eval_batch_step : [0, 1000]
[2025/02/14 20:10:35] ppocr INFO: infer_img : None
[2025/02/14 20:10:35] ppocr INFO: infer_visual_type : RU
[2025/02/14 20:10:35] ppocr INFO: log_smooth_window : 20
[2025/02/14 20:10:35] ppocr INFO: max_text_length : 50
[2025/02/14 20:10:35] ppocr INFO: max_text_nums : 30
[2025/02/14 20:10:35] ppocr INFO: pretrained_model : None
[2025/02/14 20:10:35] ppocr INFO: print_batch_step : 10
[2025/02/14 20:10:35] ppocr INFO: save_epoch_step : 10
[2025/02/14 20:10:35] ppocr INFO: save_inference_dir : None
[2025/02/14 20:10:35] ppocr INFO: save_model_dir : ./output/pgnet_r50_vd_totaltext/
[2025/02/14 20:10:35] ppocr INFO: save_res_path : ./output/pgnet_r50_vd_totaltext/predicts_pgnet.txt
[2025/02/14 20:10:35] ppocr INFO: tcl_len : 64
[2025/02/14 20:10:35] ppocr INFO: use_gpu : True
[2025/02/14 20:10:35] ppocr INFO: use_visualdl : False
[2025/02/14 20:10:35] ppocr INFO: valid_set : totaltext
[2025/02/14 20:10:35] ppocr INFO: Loss :
[2025/02/14 20:10:35] ppocr INFO: max_text_length : 50
[2025/02/14 20:10:35] ppocr INFO: max_text_nums : 30
[2025/02/14 20:10:35] ppocr INFO: name : PGLoss
[2025/02/14 20:10:35] ppocr INFO: pad_num : 36
[2025/02/14 20:10:35] ppocr INFO: tcl_bs : 64
[2025/02/14 20:10:35] ppocr INFO: Metric :
[2025/02/14 20:10:35] ppocr INFO: character_dict_path : ppocr/utils/dict/cyrillic_dict.txt
[2025/02/14 20:10:35] ppocr INFO: gt_mat_dir : ./train_data/total_text/gt
[2025/02/14 20:10:35] ppocr INFO: main_indicator : f_score_e2e
[2025/02/14 20:10:35] ppocr INFO: mode : A
[2025/02/14 20:10:35] ppocr INFO: name : E2EMetric
[2025/02/14 20:10:35] ppocr INFO: Optimizer :
[2025/02/14 20:10:35] ppocr INFO: beta1 : 0.9
[2025/02/14 20:10:35] ppocr INFO: beta2 : 0.999
[2025/02/14 20:10:35] ppocr INFO: lr :
[2025/02/14 20:10:35] ppocr INFO: learning_rate : 0.001
[2025/02/14 20:10:35] ppocr INFO: name : Cosine
[2025/02/14 20:10:35] ppocr INFO: warmup_epoch : 50
[2025/02/14 20:10:35] ppocr INFO: name : Adam
[2025/02/14 20:10:35] ppocr INFO: regularizer :
[2025/02/14 20:10:35] ppocr INFO: factor : 0.0001
[2025/02/14 20:10:35] ppocr INFO: name : L2
[2025/02/14 20:10:35] ppocr INFO: PostProcess :
[2025/02/14 20:10:35] ppocr INFO: mode : fast
[2025/02/14 20:10:35] ppocr INFO: name : PGPostProcess
[2025/02/14 20:10:35] ppocr INFO: point_gather_mode : align
[2025/02/14 20:10:35] ppocr INFO: score_thresh : 0.5
[2025/02/14 20:10:35] ppocr INFO: Train :
[2025/02/14 20:10:35] ppocr INFO: dataset :
[2025/02/14 20:10:35] ppocr INFO: data_dir : ./PP-OCRv4-dataset/train/
[2025/02/14 20:10:35] ppocr INFO: label_file_list : ['annotations/.json']
[2025/02/14 20:10:35] ppocr INFO: name : SimpleDataSet
[2025/02/14 20:10:35] ppocr INFO: ratio_list : [1.0]
[2025/02/14 20:10:35] ppocr INFO: transforms :
[2025/02/14 20:10:35] ppocr INFO: DecodeImage :
[2025/02/14 20:10:35] ppocr INFO: channel_first : False
[2025/02/14 20:10:35] ppocr INFO: img_mode : BGR
[2025/02/14 20:10:35] ppocr INFO: E2ELabelEncodeTrain : None
[2025/02/14 20:10:35] ppocr INFO: PGProcessTrain :
[2025/02/14 20:10:35] ppocr INFO: batch_size : 14
[2025/02/14 20:10:35] ppocr INFO: max_text_size : 512
[2025/02/14 20:10:35] ppocr INFO: min_crop_size : 24
[2025/02/14 20:10:35] ppocr INFO: min_text_size : 4
[2025/02/14 20:10:35] ppocr INFO: point_gather_mode : align
[2025/02/14 20:10:35] ppocr INFO: use_random_crop : False
[2025/02/14 20:10:35] ppocr INFO: use_resize : True
[2025/02/14 20:10:35] ppocr INFO: KeepKeys :
[2025/02/14 20:10:35] ppocr INFO: keep_keys : ['images', 'tcl_maps', 'tcl_label_maps', 'border_maps', 'direction_maps', 'training_masks', 'label_list', 'pos_list', 'pos_mask']
[2025/02/14 20:10:35] ppocr INFO: loader :
[2025/02/14 20:10:35] ppocr INFO: batch_size_per_card : 14
[2025/02/14 20:10:35] ppocr INFO: drop_last : True
[2025/02/14 20:10:35] ppocr INFO: num_workers : 16
[2025/02/14 20:10:35] ppocr INFO: shuffle : True
[2025/02/14 20:10:35] ppocr INFO: profiler_options : None
[2025/02/14 20:10:35] ppocr INFO: train with paddle 2.5.0 and device Place(gpu:0)
[2025/02/14 20:10:35] ppocr INFO: Initialize indexs of datasets: ['annotations/.json']
[2025/02/14 20:10:35] ppocr INFO: Using ratio list: [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1
.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0,
1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
[2025/02/14 20:12:56] ppocr INFO: Initialize indexs of datasets: ['annotations/.json']
[2025/02/14 20:12:56] ppocr INFO: Using ratio list: [1.0, 1.0, 1.0, 1.0, 1.0]
W0214 20:13:07.647279 11556 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.6, Driver API Version: 12.4, Runtime API Version: 11.2
W0214 20:13:07.656876 11556 gpu_resources.cc:149] device: 0, cuDNN Version: 8.2.
[2025/02/14 20:13:09] ppocr INFO: train dataloader has 60863 iters
[2025/02/14 20:13:09] ppocr INFO: valid dataloader has 48609 iters
[2025/02/14 20:13:09] ppocr INFO: train from scratch
[2025/02/14 20:13:09] ppocr INFO: During the training process, after the 0th iteration, an evaluation is run every 1000 iterations
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/0_0/lhr_0_1031_01222220.jpg', 'annotation
s': [{'polygon': [0.6309315762778586, 0.112991563975811, 0.9336465525072675, 0.112991563975811, 0.9336465525072675, 0.35180094838142395, 0.6309315762
778586, 0.35180094838142395], 'label': 'выругаться:', 'type': 'printed'}, {'polygon': [0.2567337627558745, 0.09653285890817642, 0.5012350392896076, 0
.09653285890817642, 0.5012350392896076, 0.3742563724517822, 0.2567337627558745, 0.3742563724517822], 'label': 'порыжеть 52772 аська клин', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_4/lhr_14_105_00125080.jpg', 'annotation
s': [{'polygon': [0.7503180503845215, 0.16061107317606607, 0.9845978021621704, 0.16061107317606607, 0.9845978021621704, 0.2652724186579386, 0.7503180503845215, 0.2652724186579386], 'label': 'Пудовкин', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/0_11/lhr_0_863_01022641.jpg', 'annotation
s': [{'polygon': [0.6674890887492881, 0.3044753074645996, 0.8171328325206435, 0.3044753074645996, 0.8171328325206435, 0.4398674964904785, 0.667489088
7492881, 0.4398674964904785], 'label': '33392', 'type': 'printed'}, {'polygon': [0.1588311510368469, 0.6093335151672363, 0.4043034095155894, 0.609333
5151672363, 0.4043034095155894, 0.8614568710327148, 0.1588311510368469, 0.8614568710327148], 'label': 'Опарина', 'type': 'printed'}, {'polygon': [0.4
107921834958713, 0.11213694512844086, 0.8656021152922124, 0.11213694512844086, 0.8656021152922124, 0.1796937733888626, 0.4107921834958713, 0.1796937733888626], 'label': 'механик-водитель убито', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_11/lhr_14_583_00691978.jpg', 'annotatio
ns': [{'polygon': [0.46629227299393317, 0.13498879969120026, 0.6020812317565247, 0.13498879969120026, 0.6020812317565247, 0.20145283639431, 0.4662922
7299393317, 0.20145283639431], 'label': 'завал(', 'type': 'printed'}, {'polygon': [0.3372459998497596, 0.8440180420875549, 0.47964564732142856, 0.844
0180420875549, 0.47964564732142856, 0.9436938166618347, 0.3372459998497596, 0.9436938166618347], 'label': 'мимо', 'type': 'printed'}, {'polygon': [0.
13667473985161974, 0.058924391865730286, 0.3344984117445055, 0.058924391865730286, 0.3344984117445055, 0.2594873905181885, 0.13667473985161974, 0.259
4873905181885], 'label': 'паевой внучка', 'type': 'printed'}, {'polygon': [0.8113468156192766, 0.23607280850410461, 0.9859493940304487, 0.23607280850
410461, 0.9859493940304487, 0.32600390911102295, 0.8113468156192766, 0.32600390911102295], 'label': 'толь пуща?', 'type': 'printed'}, {'polygon': [0.
47886332417582417, 0.4006194472312927, 0.6169260996165293, 0.4006194472312927, 0.6169260996165293, 0.5093043446540833, 0.47886332417582417, 0.5093043446540833], 'label': 'сбор', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/0_6/lhr_0_483_00572287.jpg', 'annotations
': [{'polygon': [0.7631299495697021, 0.014552081003785133, 0.946108897527059, 0.014552081003785133, 0.946108897527059, 0.16532450914382935, 0.7631299
495697021, 0.16532450914382935], 'label': 'опала-', 'type': 'printed'}, {'polygon': [0.8193318843841553, 0.8379898071289062, 0.9338462352752686, 0.83
79898071289062, 0.9338462352752686, 0.9534388780593872, 0.8193318843841553, 0.9534388780593872], 'label': 'угор', 'type': 'printed'}, {'polygon': [0.
04705489675203959, 0.15078163146972656, 0.45191915829976403, 0.15078163146972656, 0.45191915829976403, 0.20002660155296326, 0.04705489675203959, 0.20002660155296326], 'label': 'запланировать', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_17/lhr_15_119_00151731.jpg', 'annotatio
ns': [{'polygon': [0.3565429449081421, 0.7976653083922371, 0.8192929029464722, 0.7976653083922371, 0.8192929029464722, 0.9405352880084326, 0.3565429449081421, 0.9405352880084326], 'label': 'завыть 9453424154 сходни Матюшенко неослабевающий', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_9/lhr_14_451_00535562.jpg', 'annotation
s': [{'polygon': [0.10880836844444275, 0.2425532341003418, 0.4100768268108368, 0.2425532341003418, 0.4100768268108368, 0.29829275608062744, 0.10880836844444275, 0.29829275608062744], 'label': 'проторить', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/2_4/lhr_12_19_00024263.jpg', 'annotations
': [{'polygon': [0.7020158767700195, 0.6111408869425455, 0.8898781538009644, 0.6111408869425455, 0.8898781538009644, 0.6923139890034993, 0.7020158767
700195, 0.6923139890034993], 'label': 'риза', 'type': 'printed'}, {'polygon': [0.07221144437789917, 0.7680433591206869, 0.24062815308570862, 0.7680433591206869, 0.24062815308570862, 0.8497076034545898, 0.07221144437789917, 0.8497076034545898], 'label': 'баул', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/1_13/lhr_11_795_00960218.jpg', 'annotatio
ns': [{'polygon': [0.13096123933792114, 0.6723529696464539, 0.2555067737897237, 0.6723529696464539, 0.2555067737897237, 0.7424202561378479, 0.1309612
3933792114, 0.7424202561378479], 'label': '75597', 'type': 'printed'}, {'polygon': [0.7401151657104492, 0.13535887002944946, 0.8865143458048502, 0.13
535887002944946, 0.8865143458048502, 0.29110050201416016, 0.7401151657104492, 0.29110050201416016], 'label': '35667', 'type': 'printed'}, {'polygon':
[0.5721571445465088, 0.1284312903881073, 0.685986598332723, 0.1284312903881073, 0.685986598332723, 0.226767897605896, 0.5721571445465088, 0.226767897605896], 'label': '69914', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10]
作为初学者，对PP-v4的数据处理逻辑尚不熟悉，若问题表述不够清晰，还请海涵！
这个项目已经卡主很久了，真心期待您的专业解答，感谢！

[GreatV]

你的 End-to-End 训练报错可能与数据格式、数据读取逻辑或 PaddleOCR 代码适配有关。以下是一些可能的原因以及对应的解决方案：

1. 确保 JSON 标注格式正确

你的 JSON 标注格式如下：

{
  "image": "images/0_0/lhr_0_1000_01184616.jpg",
  "annotations": [
    {
      "polygon": [0.58, 0.72, 0.86, 0.72, 0.86, 0.96, 0.58, 0.96],
      "label": "воссоздать",
      "type": "printed"
    },
    {
      "polygon": [0.12, 0.35, 0.46, 0.35, 0.46, 0.62, 0.12, 0.62],
      "label": "коммунизм",
      "type": "handwriting"
    }
  ]
}

请确保：

• polygon 是 8 个数值（即 4 个点，每个点 2 个坐标）。
• label 是 字符串，不能为 null。
• type 字段不会影响训练（因为 PGNet 只关注 polygon 和 label）。

2. 检查 SimpleDataSet 实现是否正确

你修改的 SimpleDataSet 代码可能存在数据读取错误：

• 问题 1：你的 get_image_info_list 方法使用 random.sample 进行数据采样，这可能导致部分样本缺失。
  解决方案：移除 random.sample，让所有数据都被加载。
• 问题 2：你使用 encoding="windows-1251" 读取 JSON 文件，可能导致编码问题。
  解决方案：尝试 encoding="utf-8" 读取 JSON，或手动转换编码。

3. 确保 e2e_r50_vd_pg.yml 配置文件正确

你修改的配置文件可能存在以下问题：

• 问题 1：label_file_list: ["annotations/*.json"] 可能无法正确解析 JSON 文件路径。
  建议改为：

  label_file_list: ["./PP-OCRv4-dataset/train/annotations/*.json"]

• 问题 2：character_dict_path 指定了 ppocr/utils/dict/cyrillic_dict.txt，确保该字典文件存在，并且格式正确。

4. 确保 E2ELabelEncodeTrain 处理 JSON 数据正确

你的 PGNet 训练流程需要 E2ELabelEncodeTrain 进行数据预处理：

• 你可以在 SimpleDataSet 中添加 print(data)，查看是否数据正确传递到 E2ELabelEncodeTrain 处理。

5. 检查报错信息

你提供了 [训练报错信息.txt]，但没有具体内容。建议：

• 运行 python tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml 后 完整复制报错信息，以便更精确定位问题。

6. 进一步 Debug

如果仍然报错，你可以：

• 运行 python tools/train.py ... --debug 查看详细日志。
• 在 SimpleDataSet __getitem__ 方法中 print(data)，确保数据格式正确。

---

总结

• 检查 JSON 格式，确保 polygon 和 label 结构正确。
• 调整 SimpleDataSet，确保数据加载完整，不丢失样本。
• 修改 e2e_r50_vd_pg.yml，确保 label_file_list 和 character_dict_path 正确。
• 检查 E2ELabelEncodeTrain，确保数据正确传递到训练流程。
• 查看完整的报错信息，更精确地定位问题。

如果有更详细的报错信息，请提供，我们可以进一步分析！🚀

---

Response generated by feifei-bot | chatgpt-4o-latest

[Jerry200404]

报错信息PS D:\PP-OCR\PaddleOCR> python tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml
[2025/02/14 20:10:35] ppocr INFO: Architecture :
[2025/02/14 20:10:35] ppocr INFO: Backbone :
[2025/02/14 20:10:35] ppocr INFO: layers : 50
[2025/02/14 20:10:35] ppocr INFO: name : ResNet
[2025/02/14 20:10:35] ppocr INFO: Head :
[2025/02/14 20:10:35] ppocr INFO: character_dict_path : ppocr/utils/dict/cyrillic_dict.txt
[2025/02/14 20:10:35] ppocr INFO: name : PGHead
[2025/02/14 20:10:35] ppocr INFO: Neck :
[2025/02/14 20:10:35] ppocr INFO: name : PGFPN
[2025/02/14 20:10:35] ppocr INFO: Transform : None
[2025/02/14 20:10:35] ppocr INFO: algorithm : PGNet
[2025/02/14 20:10:35] ppocr INFO: model_type : e2e
[2025/02/14 20:10:35] ppocr INFO: Eval :
[2025/02/14 20:10:35] ppocr INFO: dataset :
[2025/02/14 20:10:35] ppocr INFO: data_dir : ./PP-OCRv4-dataset/test/
[2025/02/14 20:10:35] ppocr INFO: label_file_list : ['annotations/.json']
[2025/02/14 20:10:35] ppocr INFO: name : SimpleDataSet
[2025/02/14 20:10:35] ppocr INFO: transforms :
[2025/02/14 20:10:35] ppocr INFO: DecodeImage :
[2025/02/14 20:10:35] ppocr INFO: channel_first : False
[2025/02/14 20:10:35] ppocr INFO: img_mode : BGR
[2025/02/14 20:10:35] ppocr INFO: E2ELabelEncodeTest : None
[2025/02/14 20:10:35] ppocr INFO: E2EResizeForTest :
[2025/02/14 20:10:35] ppocr INFO: max_side_len : 768
[2025/02/14 20:10:35] ppocr INFO: NormalizeImage :
[2025/02/14 20:10:35] ppocr INFO: mean : [0.485, 0.456, 0.406]
[2025/02/14 20:10:35] ppocr INFO: order : hwc
[2025/02/14 20:10:35] ppocr INFO: scale : 1./255.
[2025/02/14 20:10:35] ppocr INFO: std : [0.229, 0.224, 0.225]
[2025/02/14 20:10:35] ppocr INFO: ToCHWImage : None
[2025/02/14 20:10:35] ppocr INFO: KeepKeys :
[2025/02/14 20:10:35] ppocr INFO: keep_keys : ['image', 'shape', 'polys', 'texts', 'ignore_tags', 'img_id']
[2025/02/14 20:10:35] ppocr INFO: loader :
[2025/02/14 20:10:35] ppocr INFO: batch_size_per_card : 1
[2025/02/14 20:10:35] ppocr INFO: drop_last : False
[2025/02/14 20:10:35] ppocr INFO: num_workers : 2
[2025/02/14 20:10:35] ppocr INFO: shuffle : False
[2025/02/14 20:10:35] ppocr INFO: Global :
[2025/02/14 20:10:35] ppocr INFO: cal_metric_during_train : False
[2025/02/14 20:10:35] ppocr INFO: character_dict_path : ppocr/utils/dict/cyrillic_dict.txt
[2025/02/14 20:10:35] ppocr INFO: character_type : RU
[2025/02/14 20:10:35] ppocr INFO: checkpoints : None
[2025/02/14 20:10:35] ppocr INFO: distributed : False
[2025/02/14 20:10:35] ppocr INFO: epoch_num : 600
[2025/02/14 20:10:35] ppocr INFO: eval_batch_step : [0, 1000]
[2025/02/14 20:10:35] ppocr INFO: infer_img : None
[2025/02/14 20:10:35] ppocr INFO: infer_visual_type : RU
[2025/02/14 20:10:35] ppocr INFO: log_smooth_window : 20
[2025/02/14 20:10:35] ppocr INFO: max_text_length : 50
[2025/02/14 20:10:35] ppocr INFO: max_text_nums : 30
[2025/02/14 20:10:35] ppocr INFO: pretrained_model : None
[2025/02/14 20:10:35] ppocr INFO: print_batch_step : 10
[2025/02/14 20:10:35] ppocr INFO: save_epoch_step : 10
[2025/02/14 20:10:35] ppocr INFO: save_inference_dir : None
[2025/02/14 20:10:35] ppocr INFO: save_model_dir : ./output/pgnet_r50_vd_totaltext/
[2025/02/14 20:10:35] ppocr INFO: save_res_path : ./output/pgnet_r50_vd_totaltext/predicts_pgnet.txt
[2025/02/14 20:10:35] ppocr INFO: tcl_len : 64
[2025/02/14 20:10:35] ppocr INFO: use_gpu : True
[2025/02/14 20:10:35] ppocr INFO: use_visualdl : False
[2025/02/14 20:10:35] ppocr INFO: valid_set : totaltext
[2025/02/14 20:10:35] ppocr INFO: Loss :
[2025/02/14 20:10:35] ppocr INFO: max_text_length : 50
[2025/02/14 20:10:35] ppocr INFO: max_text_nums : 30
[2025/02/14 20:10:35] ppocr INFO: name : PGLoss
[2025/02/14 20:10:35] ppocr INFO: pad_num : 36
[2025/02/14 20:10:35] ppocr INFO: tcl_bs : 64
[2025/02/14 20:10:35] ppocr INFO: Metric :
[2025/02/14 20:10:35] ppocr INFO: character_dict_path : ppocr/utils/dict/cyrillic_dict.txt
[2025/02/14 20:10:35] ppocr INFO: gt_mat_dir : ./train_data/total_text/gt
[2025/02/14 20:10:35] ppocr INFO: main_indicator : f_score_e2e
[2025/02/14 20:10:35] ppocr INFO: mode : A
[2025/02/14 20:10:35] ppocr INFO: name : E2EMetric
[2025/02/14 20:10:35] ppocr INFO: Optimizer :
[2025/02/14 20:10:35] ppocr INFO: beta1 : 0.9
[2025/02/14 20:10:35] ppocr INFO: beta2 : 0.999
[2025/02/14 20:10:35] ppocr INFO: lr :
[2025/02/14 20:10:35] ppocr INFO: learning_rate : 0.001
[2025/02/14 20:10:35] ppocr INFO: name : Cosine
[2025/02/14 20:10:35] ppocr INFO: warmup_epoch : 50
[2025/02/14 20:10:35] ppocr INFO: name : Adam
[2025/02/14 20:10:35] ppocr INFO: regularizer :
[2025/02/14 20:10:35] ppocr INFO: factor : 0.0001
[2025/02/14 20:10:35] ppocr INFO: name : L2
[2025/02/14 20:10:35] ppocr INFO: PostProcess :
[2025/02/14 20:10:35] ppocr INFO: mode : fast
[2025/02/14 20:10:35] ppocr INFO: name : PGPostProcess
[2025/02/14 20:10:35] ppocr INFO: point_gather_mode : align
[2025/02/14 20:10:35] ppocr INFO: score_thresh : 0.5
[2025/02/14 20:10:35] ppocr INFO: Train :
[2025/02/14 20:10:35] ppocr INFO: dataset :
[2025/02/14 20:10:35] ppocr INFO: data_dir : ./PP-OCRv4-dataset/train/
[2025/02/14 20:10:35] ppocr INFO: label_file_list : ['annotations/.json']
[2025/02/14 20:10:35] ppocr INFO: name : SimpleDataSet
[2025/02/14 20:10:35] ppocr INFO: ratio_list : [1.0]
[2025/02/14 20:10:35] ppocr INFO: transforms :
[2025/02/14 20:10:35] ppocr INFO: DecodeImage :
[2025/02/14 20:10:35] ppocr INFO: channel_first : False
[2025/02/14 20:10:35] ppocr INFO: img_mode : BGR
[2025/02/14 20:10:35] ppocr INFO: E2ELabelEncodeTrain : None
[2025/02/14 20:10:35] ppocr INFO: PGProcessTrain :
[2025/02/14 20:10:35] ppocr INFO: batch_size : 14
[2025/02/14 20:10:35] ppocr INFO: max_text_size : 512
[2025/02/14 20:10:35] ppocr INFO: min_crop_size : 24
[2025/02/14 20:10:35] ppocr INFO: min_text_size : 4
[2025/02/14 20:10:35] ppocr INFO: point_gather_mode : align
[2025/02/14 20:10:35] ppocr INFO: use_random_crop : False
[2025/02/14 20:10:35] ppocr INFO: use_resize : True
[2025/02/14 20:10:35] ppocr INFO: KeepKeys :
[2025/02/14 20:10:35] ppocr INFO: keep_keys : ['images', 'tcl_maps', 'tcl_label_maps', 'border_maps', 'direction_maps', 'training_masks', 'label_list', 'pos_list', 'pos_mask']
[2025/02/14 20:10:35] ppocr INFO: loader :
[2025/02/14 20:10:35] ppocr INFO: batch_size_per_card : 14
[2025/02/14 20:10:35] ppocr INFO: drop_last : True
[2025/02/14 20:10:35] ppocr INFO: num_workers : 16
[2025/02/14 20:10:35] ppocr INFO: shuffle : True
[2025/02/14 20:10:35] ppocr INFO: profiler_options : None
[2025/02/14 20:10:35] ppocr INFO: train with paddle 2.5.0 and device Place(gpu:0)
[2025/02/14 20:10:35] ppocr INFO: Initialize indexs of datasets: ['annotations/.json']
[2025/02/14 20:10:35] ppocr INFO: Using ratio list: [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1
.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0,
1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
[2025/02/14 20:12:56] ppocr INFO: Initialize indexs of datasets: ['annotations/.json']
[2025/02/14 20:12:56] ppocr INFO: Using ratio list: [1.0, 1.0, 1.0, 1.0, 1.0]
W0214 20:13:07.647279 11556 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.6, Driver API Version: 12.4, Runtime API Version: 11.2
W0214 20:13:07.656876 11556 gpu_resources.cc:149] device: 0, cuDNN Version: 8.2.
[2025/02/14 20:13:09] ppocr INFO: train dataloader has 60863 iters
[2025/02/14 20:13:09] ppocr INFO: valid dataloader has 48609 iters
[2025/02/14 20:13:09] ppocr INFO: train from scratch
[2025/02/14 20:13:09] ppocr INFO: During the training process, after the 0th iteration, an evaluation is run every 1000 iterations
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/0_0/lhr_0_1031_01222220.jpg', 'annotation
s': [{'polygon': [0.6309315762778586, 0.112991563975811, 0.9336465525072675, 0.112991563975811, 0.9336465525072675, 0.35180094838142395, 0.6309315762
778586, 0.35180094838142395], 'label': 'выругаться:', 'type': 'printed'}, {'polygon': [0.2567337627558745, 0.09653285890817642, 0.5012350392896076, 0
.09653285890817642, 0.5012350392896076, 0.3742563724517822, 0.2567337627558745, 0.3742563724517822], 'label': 'порыжеть 52772 аська клин', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_4/lhr_14_105_00125080.jpg', 'annotation
s': [{'polygon': [0.7503180503845215, 0.16061107317606607, 0.9845978021621704, 0.16061107317606607, 0.9845978021621704, 0.2652724186579386, 0.7503180503845215, 0.2652724186579386], 'label': 'Пудовкин', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/0_11/lhr_0_863_01022641.jpg', 'annotation
s': [{'polygon': [0.6674890887492881, 0.3044753074645996, 0.8171328325206435, 0.3044753074645996, 0.8171328325206435, 0.4398674964904785, 0.667489088
7492881, 0.4398674964904785], 'label': '33392', 'type': 'printed'}, {'polygon': [0.1588311510368469, 0.6093335151672363, 0.4043034095155894, 0.609333
5151672363, 0.4043034095155894, 0.8614568710327148, 0.1588311510368469, 0.8614568710327148], 'label': 'Опарина', 'type': 'printed'}, {'polygon': [0.4
107921834958713, 0.11213694512844086, 0.8656021152922124, 0.11213694512844086, 0.8656021152922124, 0.1796937733888626, 0.4107921834958713, 0.1796937733888626], 'label': 'механик-водитель убито', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_11/lhr_14_583_00691978.jpg', 'annotatio
ns': [{'polygon': [0.46629227299393317, 0.13498879969120026, 0.6020812317565247, 0.13498879969120026, 0.6020812317565247, 0.20145283639431, 0.4662922
7299393317, 0.20145283639431], 'label': 'завал(', 'type': 'printed'}, {'polygon': [0.3372459998497596, 0.8440180420875549, 0.47964564732142856, 0.844
0180420875549, 0.47964564732142856, 0.9436938166618347, 0.3372459998497596, 0.9436938166618347], 'label': 'мимо', 'type': 'printed'}, {'polygon': [0.
13667473985161974, 0.058924391865730286, 0.3344984117445055, 0.058924391865730286, 0.3344984117445055, 0.2594873905181885, 0.13667473985161974, 0.259
4873905181885], 'label': 'паевой внучка', 'type': 'printed'}, {'polygon': [0.8113468156192766, 0.23607280850410461, 0.9859493940304487, 0.23607280850
410461, 0.9859493940304487, 0.32600390911102295, 0.8113468156192766, 0.32600390911102295], 'label': 'толь пуща?', 'type': 'printed'}, {'polygon': [0.
47886332417582417, 0.4006194472312927, 0.6169260996165293, 0.4006194472312927, 0.6169260996165293, 0.5093043446540833, 0.47886332417582417, 0.5093043446540833], 'label': 'сбор', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/0_6/lhr_0_483_00572287.jpg', 'annotations
': [{'polygon': [0.7631299495697021, 0.014552081003785133, 0.946108897527059, 0.014552081003785133, 0.946108897527059, 0.16532450914382935, 0.7631299
495697021, 0.16532450914382935], 'label': 'опала-', 'type': 'printed'}, {'polygon': [0.8193318843841553, 0.8379898071289062, 0.9338462352752686, 0.83
79898071289062, 0.9338462352752686, 0.9534388780593872, 0.8193318843841553, 0.9534388780593872], 'label': 'угор', 'type': 'printed'}, {'polygon': [0.
04705489675203959, 0.15078163146972656, 0.45191915829976403, 0.15078163146972656, 0.45191915829976403, 0.20002660155296326, 0.04705489675203959, 0.20002660155296326], 'label': 'запланировать', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_17/lhr_15_119_00151731.jpg', 'annotatio
ns': [{'polygon': [0.3565429449081421, 0.7976653083922371, 0.8192929029464722, 0.7976653083922371, 0.8192929029464722, 0.9405352880084326, 0.3565429449081421, 0.9405352880084326], 'label': 'завыть 9453424154 сходни Матюшенко неослабевающий', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/3_9/lhr_14_451_00535562.jpg', 'annotation
s': [{'polygon': [0.10880836844444275, 0.2425532341003418, 0.4100768268108368, 0.2425532341003418, 0.4100768268108368, 0.29829275608062744, 0.10880836844444275, 0.29829275608062744], 'label': 'проторить', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/2_4/lhr_12_19_00024263.jpg', 'annotations
': [{'polygon': [0.7020158767700195, 0.6111408869425455, 0.8898781538009644, 0.6111408869425455, 0.8898781538009644, 0.6923139890034993, 0.7020158767
700195, 0.6923139890034993], 'label': 'риза', 'type': 'printed'}, {'polygon': [0.07221144437789917, 0.7680433591206869, 0.24062815308570862, 0.7680433591206869, 0.24062815308570862, 0.8497076034545898, 0.07221144437789917, 0.8497076034545898], 'label': 'баул', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10] ppocr ERROR: Error occurred while parsing {'image': './PP-OCRv4-dataset/train/images/1_13/lhr_11_795_00960218.jpg', 'annotatio
ns': [{'polygon': [0.13096123933792114, 0.6723529696464539, 0.2555067737897237, 0.6723529696464539, 0.2555067737897237, 0.7424202561378479, 0.1309612
3933792114, 0.7424202561378479], 'label': '75597', 'type': 'printed'}, {'polygon': [0.7401151657104492, 0.13535887002944946, 0.8865143458048502, 0.13
535887002944946, 0.8865143458048502, 0.29110050201416016, 0.7401151657104492, 0.29110050201416016], 'label': '35667', 'type': 'printed'}, {'polygon':
[0.5721571445465088, 0.1284312903881073, 0.685986598332723, 0.1284312903881073, 0.685986598332723, 0.226767897605896, 0.5721571445465088, 0.226767897605896], 'label': '69914', 'type': 'printed'}]}: 'label'
[2025/02/14 20:13:10]