I am trying to train rec model - en_PP-OCRv3_rec_train but i am getting an AssertionError: invalid input 'img' in DecodeImage #12527

kishore-work-hard · 2024-05-28T21:46:20Z

kishore-work-hard
May 28, 2024

i am using an ubuntu machine. and i have installed all requirements to run and train paddle OCR. I have modified the config file configs/rec/PP-OCRv3/en_PP-OCRv3_rec.yml (i just set the training image and label path. testing image and label path)

Global:
  debug: false
  use_gpu: true
  epoch_num: 500
  log_smooth_window: 20
  print_batch_step: 10
  save_model_dir: ./output/v3_en_mobile
  save_epoch_step: 3
  eval_batch_step: [0, 2000]
  cal_metric_during_train: true
  pretrained_model:
  checkpoints:
  save_inference_dir:
  use_visualdl: false
  infer_img: doc/imgs_words/ch/word_1.jpg
  character_dict_path: ppocr/utils/en_dict.txt
  max_text_length: &max_text_length 25
  infer_mode: false
  use_space_char: true
  distributed: true
  save_res_path: ./output/rec/predicts_ppocrv3_en.txt


Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.999
  lr:
    name: Cosine
    learning_rate: 0.001
    warmup_epoch: 5
  regularizer:
    name: L2
    factor: 3.0e-05


Architecture:
  model_type: rec
  algorithm: SVTR_LCNet
  Transform:
  Backbone:
    name: MobileNetV1Enhance
    scale: 0.5
    last_conv_stride: [1, 2]
    last_pool_type: avg
    last_pool_kernel_size: [2, 2]
  Head:
    name: MultiHead
    head_list:
      - CTCHead:
          Neck:
            name: svtr
            dims: 64
            depth: 2
            hidden_dims: 120
            use_guide: True
          Head:
            fc_decay: 0.00001
      - SARHead:
          enc_dim: 512
          max_text_length: *max_text_length

Loss:
  name: MultiLoss
  loss_config_list:
    - CTCLoss:
    - SARLoss:

PostProcess:  
  name: CTCLabelDecode

Metric:
  name: RecMetric
  main_indicator: acc
  ignore_space: False

Train:
  dataset:
    name: SimpleDataSet
    data_dir: ../../../data/scut_data/scut_train
    ext_op_transform_idx: 1
    label_file_list:
    - ../../../data/scut_data/tr.txt
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - RecConAug:
        prob: 0.5
        ext_data_num: 2
        image_shape: [48, 320, 3]
        max_text_length: *max_text_length
    - RecAug:
    - MultiLabelEncode:
    - RecResizeImg:
        image_shape: [3, 48, 320]
    - KeepKeys:
        keep_keys:
        - image
        - label_ctc
        - label_sar
        - length
        - valid_ratio
  loader:
    shuffle: true
    batch_size_per_card: 128
    drop_last: true
    num_workers: 4
Eval:
  dataset:
    name: SimpleDataSet
    data_dir: ../../../data/scut_data/scut_test
    label_file_list:
    - ../../../data/scut_data/tt.txt
    transforms:
    - DecodeImage:
        img_mode: BGR
        channel_first: false
    - MultiLabelEncode:
    - RecResizeImg:
        image_shape: [3, 48, 320]
    - KeepKeys:
        keep_keys:
        - image
        - label_ctc
        - label_sar
        - length
        - valid_ratio
  loader:
    shuffle: false
    drop_last: false
    batch_size_per_card: 128
    num_workers: 4

i am using scut dataset which i downloaded from online. which contains .txt file annotation in this format - filename tab text

006085.jpg	more
006086.jpg	Honey Cre
006087.jpg	HOney Crem
006088.jpg	eam p
006089.jpg	emium soft

i have done the dos2unix conversions for these files since they are copied from windows pc.
i am also using the dictionary file ppocr/utils/EN_symbol_dict.txt

when i run the script -
python tools/train.py -c configs/rec/PP-OCRv3/en_PP-OCRv3_rec.yml -o Global.pretrained_model=pretrain_models/en_PP-OCRv3_rec_train/best_accuracy

i am getting this error -

[2024/05/29 03:10:34] ppocr INFO: epoch: [2/500], global_step: 60, lr: 0.000211, acc: 0.000000, norm_edit_dis: 0.055315, CTCLoss: 130.410622, SARLoss: 3.701646, loss: 134.093987, avg_reader_cost: 0.26473 s, avg_batch_cost: 0.66573 s, avg_samples: 128.0, ips: 192.27054 samples/s, eta: 4:12:04, max_mem_reserved: 10218 MB, max_mem_allocated: 10011 MB
[2024/05/29 03:10:35] ppocr ERROR: When parsing line 000259.jpg Uppingham
, error happened with msg: Traceback (most recent call last):
  File "/home/elvicto/.ghostshell/Test/train_text/PaddleOCR/ppocr/data/simple_dataset.py", line 158, in __getitem__
    data['ext_data'] = self.get_ext_data()
  File "/home/elvicto/.ghostshell/Test/train_text/PaddleOCR/ppocr/data/simple_dataset.py", line 132, in get_ext_data
    data = transform(data, load_data_ops)
  File "/home/elvicto/.ghostshell/Test/train_text/PaddleOCR/ppocr/data/imaug/__init__.py", line 56, in transform
    data = op(data)
  File "/home/elvicto/.ghostshell/Test/train_text/PaddleOCR/ppocr/data/imaug/operators.py", line 48, in __call__
    assert type(img) is bytes and len(
AssertionError: invalid input 'img' in DecodeImage

what is the issue?

Answered by kishore-work-hard

Jun 14, 2024

The issue has been resolved.
I wrote a script to remove images with length = 0 and type != <byte>

Thank you @GreatV For Your Support.
You Saved Me

View full answer

GreatV · 2024-05-29T02:14:05Z

GreatV
May 29, 2024
Maintainer

Please check whether the data reading path, label format, etc. are correct or not!

AssertionError: invalid input 'img' in DecodeImage #7195

1 reply

kishore-work-hard May 29, 2024
Author

i have provided the label format - image_name tab_space text
its a label.txt file

006085.jpg	more
006086.jpg	Honey Cre
006087.jpg	HOney Crem
006088.jpg	eam p
006089.jpg	emium soft

path is also correct only.

kishore-work-hard · 2024-06-05T15:10:24Z

kishore-work-hard
Jun 5, 2024
Author

please help anyone?

0 replies

GreatV · 2024-06-06T00:45:05Z

GreatV
Jun 6, 2024
Maintainer

Is the data directory organized like this?

tree -L 2
.
├── ocr_det
│   └── readme.md
└── ocr_rec
    ├── rec_train.txt
    ├── rec_val.txt
    ├── train
    └── val

rec_train.txt

train/img_0.jpg	汇丰晋信
train/img_1.jpg	受理时间:
train/img_2.jpg	2020.6.15
train/img_3.jpg	开放式基金账户业务申请表一机构
train/img_4.jpg	特别提示:
train/img_5.jpg	基金账号:

9 replies

kishore-work-hard Jun 12, 2024
Author

Did you really have 004807.jpg?
terminal result shell script -

this is the image file -

kishore-work-hard Jun 12, 2024
Author

That's weird, can you provide some data so we can test it?

what kind of data do u need to test it? a sample dataset ?

GreatV Jun 12, 2024
Maintainer

@kishore-work-hard yes, a sample dataset

kishore-work-hard Jun 12, 2024
Author

@kishore-work-hard yes, a sample dataset

i have attached a google drive link to a folder which contains all datas and files i used to train rec model. please go through it.
Dataset And Other Used Files

if this works for you then please guide me. i really want to learn to train a model.

kishore-work-hard Jun 12, 2024
Author

@kishore-work-hard yes, a sample dataset

i tried to do things fresh. created new folder... cloned the git repository.. and this time i noted down all the steps i have made and the corresponding results (NO change in the error)

please go through this text file and see if i made any mistakes.

my_rec_ocr_installtion_steps.txt

GreatV · 2024-06-13T02:18:27Z

GreatV
Jun 13, 2024
Maintainer

@kishore-work-hard, your data works well for me.

python tools/train.py -c configs/rec/PP-OCRv3/en_PP-OCRv3_rec.yml
[2024/06/13 02:17:17] ppocr INFO: Architecture : 
[2024/06/13 02:17:17] ppocr INFO:     Backbone : 
[2024/06/13 02:17:17] ppocr INFO:         last_conv_stride : [1, 2]
[2024/06/13 02:17:17] ppocr INFO:         last_pool_kernel_size : [2, 2]
[2024/06/13 02:17:17] ppocr INFO:         last_pool_type : avg
[2024/06/13 02:17:17] ppocr INFO:         name : MobileNetV1Enhance
[2024/06/13 02:17:17] ppocr INFO:         scale : 0.5
[2024/06/13 02:17:17] ppocr INFO:     Head : 
[2024/06/13 02:17:17] ppocr INFO:         head_list : 
[2024/06/13 02:17:17] ppocr INFO:             CTCHead : 
[2024/06/13 02:17:17] ppocr INFO:                 Head : 
[2024/06/13 02:17:17] ppocr INFO:                     fc_decay : 1e-05
[2024/06/13 02:17:17] ppocr INFO:                 Neck : 
[2024/06/13 02:17:17] ppocr INFO:                     depth : 2
[2024/06/13 02:17:17] ppocr INFO:                     dims : 64
[2024/06/13 02:17:17] ppocr INFO:                     hidden_dims : 120
[2024/06/13 02:17:17] ppocr INFO:                     name : svtr
[2024/06/13 02:17:17] ppocr INFO:                     use_guide : True
[2024/06/13 02:17:17] ppocr INFO:             SARHead : 
[2024/06/13 02:17:17] ppocr INFO:                 enc_dim : 512
[2024/06/13 02:17:17] ppocr INFO:                 max_text_length : 25
[2024/06/13 02:17:17] ppocr INFO:         name : MultiHead
[2024/06/13 02:17:17] ppocr INFO:     Transform : None
[2024/06/13 02:17:17] ppocr INFO:     algorithm : SVTR_LCNet
[2024/06/13 02:17:17] ppocr INFO:     model_type : rec
[2024/06/13 02:17:17] ppocr INFO: Eval : 
[2024/06/13 02:17:17] ppocr INFO:     dataset : 
[2024/06/13 02:17:17] ppocr INFO:         data_dir : ./train_data
[2024/06/13 02:17:17] ppocr INFO:         label_file_list : ['./train_data/scut_test.txt']
[2024/06/13 02:17:17] ppocr INFO:         name : SimpleDataSet
[2024/06/13 02:17:17] ppocr INFO:         transforms : 
[2024/06/13 02:17:17] ppocr INFO:             DecodeImage : 
[2024/06/13 02:17:17] ppocr INFO:                 channel_first : False
[2024/06/13 02:17:17] ppocr INFO:                 img_mode : BGR
[2024/06/13 02:17:17] ppocr INFO:             MultiLabelEncode : None
[2024/06/13 02:17:17] ppocr INFO:             RecResizeImg : 
[2024/06/13 02:17:17] ppocr INFO:                 image_shape : [3, 48, 320]
[2024/06/13 02:17:17] ppocr INFO:             KeepKeys : 
[2024/06/13 02:17:17] ppocr INFO:                 keep_keys : ['image', 'label_ctc', 'label_sar', 'length', 'valid_ratio']
[2024/06/13 02:17:17] ppocr INFO:     loader : 
[2024/06/13 02:17:17] ppocr INFO:         batch_size_per_card : 128
[2024/06/13 02:17:17] ppocr INFO:         drop_last : False
[2024/06/13 02:17:17] ppocr INFO:         num_workers : 4
[2024/06/13 02:17:17] ppocr INFO:         shuffle : False
[2024/06/13 02:17:17] ppocr INFO: Global : 
[2024/06/13 02:17:17] ppocr INFO:     cal_metric_during_train : True
[2024/06/13 02:17:17] ppocr INFO:     character_dict_path : ./EN_symbol_dict.txt
[2024/06/13 02:17:17] ppocr INFO:     checkpoints : None
[2024/06/13 02:17:17] ppocr INFO:     debug : False
[2024/06/13 02:17:17] ppocr INFO:     distributed : False
[2024/06/13 02:17:17] ppocr INFO:     epoch_num : 500
[2024/06/13 02:17:17] ppocr INFO:     eval_batch_step : [0, 2000]
[2024/06/13 02:17:17] ppocr INFO:     infer_img : doc/imgs_words/ch/word_1.jpg
[2024/06/13 02:17:17] ppocr INFO:     infer_mode : False
[2024/06/13 02:17:17] ppocr INFO:     log_smooth_window : 20
[2024/06/13 02:17:17] ppocr INFO:     max_text_length : 25
[2024/06/13 02:17:17] ppocr INFO:     pretrained_model : None
[2024/06/13 02:17:17] ppocr INFO:     print_batch_step : 10
[2024/06/13 02:17:17] ppocr INFO:     save_epoch_step : 3
[2024/06/13 02:17:17] ppocr INFO:     save_inference_dir : None
[2024/06/13 02:17:17] ppocr INFO:     save_model_dir : ./output/v3_en_mobile
[2024/06/13 02:17:17] ppocr INFO:     save_res_path : ./output/rec/predicts_ppocrv3_en.txt
[2024/06/13 02:17:17] ppocr INFO:     use_gpu : True
[2024/06/13 02:17:17] ppocr INFO:     use_space_char : True
[2024/06/13 02:17:17] ppocr INFO:     use_visualdl : False
[2024/06/13 02:17:17] ppocr INFO: Loss : 
[2024/06/13 02:17:17] ppocr INFO:     loss_config_list : 
[2024/06/13 02:17:17] ppocr INFO:         CTCLoss : None
[2024/06/13 02:17:17] ppocr INFO:         SARLoss : None
[2024/06/13 02:17:17] ppocr INFO:     name : MultiLoss
[2024/06/13 02:17:17] ppocr INFO: Metric : 
[2024/06/13 02:17:17] ppocr INFO:     ignore_space : False
[2024/06/13 02:17:17] ppocr INFO:     main_indicator : acc
[2024/06/13 02:17:17] ppocr INFO:     name : RecMetric
[2024/06/13 02:17:17] ppocr INFO: Optimizer : 
[2024/06/13 02:17:17] ppocr INFO:     beta1 : 0.9
[2024/06/13 02:17:17] ppocr INFO:     beta2 : 0.999
[2024/06/13 02:17:17] ppocr INFO:     lr : 
[2024/06/13 02:17:17] ppocr INFO:         learning_rate : 0.001
[2024/06/13 02:17:17] ppocr INFO:         name : Cosine
[2024/06/13 02:17:17] ppocr INFO:         warmup_epoch : 5
[2024/06/13 02:17:17] ppocr INFO:     name : Adam
[2024/06/13 02:17:17] ppocr INFO:     regularizer : 
[2024/06/13 02:17:17] ppocr INFO:         factor : 3e-05
[2024/06/13 02:17:17] ppocr INFO:         name : L2
[2024/06/13 02:17:17] ppocr INFO: PostProcess : 
[2024/06/13 02:17:17] ppocr INFO:     name : CTCLabelDecode
[2024/06/13 02:17:17] ppocr INFO: Train : 
[2024/06/13 02:17:17] ppocr INFO:     dataset : 
[2024/06/13 02:17:17] ppocr INFO:         data_dir : ./train_data/
[2024/06/13 02:17:17] ppocr INFO:         ext_op_transform_idx : 1
[2024/06/13 02:17:17] ppocr INFO:         label_file_list : ['./train_data/scut_train.txt']
[2024/06/13 02:17:17] ppocr INFO:         name : SimpleDataSet
[2024/06/13 02:17:17] ppocr INFO:         transforms : 
[2024/06/13 02:17:17] ppocr INFO:             DecodeImage : 
[2024/06/13 02:17:17] ppocr INFO:                 channel_first : False
[2024/06/13 02:17:17] ppocr INFO:                 img_mode : BGR
[2024/06/13 02:17:17] ppocr INFO:             RecConAug : 
[2024/06/13 02:17:17] ppocr INFO:                 ext_data_num : 2
[2024/06/13 02:17:17] ppocr INFO:                 image_shape : [48, 320, 3]
[2024/06/13 02:17:17] ppocr INFO:                 max_text_length : 25
[2024/06/13 02:17:17] ppocr INFO:                 prob : 0.5
[2024/06/13 02:17:17] ppocr INFO:             RecAug : None
[2024/06/13 02:17:17] ppocr INFO:             MultiLabelEncode : None
[2024/06/13 02:17:17] ppocr INFO:             RecResizeImg : 
[2024/06/13 02:17:17] ppocr INFO:                 image_shape : [3, 48, 320]
[2024/06/13 02:17:17] ppocr INFO:             KeepKeys : 
[2024/06/13 02:17:17] ppocr INFO:                 keep_keys : ['image', 'label_ctc', 'label_sar', 'length', 'valid_ratio']
[2024/06/13 02:17:17] ppocr INFO:     loader : 
[2024/06/13 02:17:17] ppocr INFO:         batch_size_per_card : 128
[2024/06/13 02:17:17] ppocr INFO:         drop_last : True
[2024/06/13 02:17:17] ppocr INFO:         num_workers : 4
[2024/06/13 02:17:17] ppocr INFO:         shuffle : True
[2024/06/13 02:17:17] ppocr INFO: profiler_options : None
[2024/06/13 02:17:17] ppocr INFO: train with paddle 2.6.1 and device Place(gpu:0)
[2024/06/13 02:17:17] ppocr INFO: Initialize indexs of datasets:['./train_data/scut_train.txt']
list index out of range
[2024/06/13 02:17:17] ppocr INFO: Initialize indexs of datasets:['./train_data/scut_test.txt']
W0613 02:17:17.444226 59468 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.9, Driver API Version: 12.4, Runtime API Version: 12.0
W0613 02:17:17.444650 59468 gpu_resources.cc:164] device: 0, cuDNN Version: 8.9.
[2024/06/13 02:17:17] ppocr INFO: train dataloader has 47 iters
[2024/06/13 02:17:17] ppocr INFO: valid dataloader has 13 iters
[2024/06/13 02:17:17] ppocr INFO: train from scratch
[2024/06/13 02:17:17] ppocr INFO: During the training process, after the 0th iteration, an evaluation is run every 2000 iterations
[2024/06/13 02:17:22] ppocr INFO: epoch: [1/500], global_step: 10, lr: 0.000019, acc: 0.000000, norm_edit_dis: 0.030794, CTCLoss: 156.278442, SARLoss: 4.659046, loss: 160.950493, avg_reader_cost: 0.38393 s, avg_batch_cost: 0.53561 s, avg_samples: 128.0, ips: 238.98138 samples/s, eta: 3:29:41, max_mem_reserved: 10185 MB, max_mem_allocated: 10018 MB
[2024/06/13 02:17:27] ppocr INFO: epoch: [1/500], global_step: 20, lr: 0.000040, acc: 0.000000, norm_edit_dis: 0.034581, CTCLoss: 154.871956, SARLoss: 4.587813, loss: 159.452431, avg_reader_cost: 0.31214 s, avg_batch_cost: 0.43427 s, avg_samples: 128.0, ips: 294.74561 samples/s, eta: 3:09:46, max_mem_reserved: 10185 MB, max_mem_allocated: 10018 MB
[2024/06/13 02:17:31] ppocr INFO: epoch: [1/500], global_step: 30, lr: 0.000083, acc: 0.000000, norm_edit_dis: 0.044091, CTCLoss: 150.188049, SARLoss: 4.391426, loss: 154.592163, avg_reader_cost: 0.30044 s, avg_batch_cost: 0.42259 s, avg_samples: 128.0, ips: 302.89135 samples/s, eta: 3:01:33, max_mem_reserved: 10185 MB, max_mem_allocated: 10018 MB
[2024/06/13 02:17:34] ppocr INFO: epoch: [1/500], global_step: 40, lr: 0.000126, acc: 0.000000

19 replies

GreatV Jun 13, 2024
Maintainer

I have no idea anymore, maybe you need to change the computer.

kishore-work-hard Jun 13, 2024
Author

It appears that OpenCV is reading an empty image, or there might be an I/O issue. What version of OpenCV are you using?
opencv-contrib-python 4.6.0.66
opencv-python         4.6.0.66

can you send me a different dataset ? so i can try with it?
maybe some images are bad in my dataset while uploading it. idk

kishore-work-hard Jun 13, 2024
Author

I have no idea anymore, maybe you need to change the computer.

brooo don't make me cry!! lmao

GreatV Jun 13, 2024
Maintainer

maybe some images are bad in my dataset while uploading it.

I think you can delete the image that causes the error or simply ignore this error message.

dataset for rec can refer:

https://github.com/PaddlePaddle/PaddleOCR/blob/main/doc/doc_en/recognition_en.md

kishore-work-hard Jun 13, 2024
Author

maybe some images are bad in my dataset while uploading it.

I think you can delete the image that causes the error or simply ignore this error message.

dataset for rec can refer:

https://github.com/PaddlePaddle/PaddleOCR/blob/main/doc/doc_en/recognition_en.md

@GreatV I found the dataset. I would like to train a model in PaddleOCRv4. is there any English document which i could refer to train that?

kishore-work-hard · 2024-06-14T02:41:29Z

kishore-work-hard
Jun 14, 2024
Author

The issue has been resolved.
I wrote a script to remove images with length = 0 and type != <byte>

Thank you @GreatV For Your Support.
You Saved Me

0 replies

I am trying to train rec model - en_PP-OCRv3_rec_train but i am getting an AssertionError: invalid input 'img' in DecodeImage #12527

Uh oh!

kishore-work-hard May 28, 2024

Replies: 5 comments · 29 replies

Uh oh!

GreatV May 29, 2024 Maintainer

Uh oh!

kishore-work-hard May 29, 2024 Author

Uh oh!

kishore-work-hard Jun 5, 2024 Author

Uh oh!

GreatV Jun 6, 2024 Maintainer

Uh oh!

kishore-work-hard Jun 12, 2024 Author

Uh oh!

kishore-work-hard Jun 12, 2024 Author

Uh oh!

GreatV Jun 12, 2024 Maintainer

Uh oh!

Uh oh!

kishore-work-hard Jun 12, 2024 Author

Uh oh!

Uh oh!

kishore-work-hard Jun 12, 2024 Author

Uh oh!

GreatV Jun 13, 2024 Maintainer

Uh oh!

GreatV Jun 13, 2024 Maintainer

Uh oh!

kishore-work-hard Jun 13, 2024 Author

Uh oh!

kishore-work-hard Jun 13, 2024 Author

Uh oh!

Uh oh!

GreatV Jun 13, 2024 Maintainer

Uh oh!

Uh oh!

kishore-work-hard Jun 13, 2024 Author

Uh oh!

Uh oh!

kishore-work-hard Jun 14, 2024 Author

kishore-work-hard
May 28, 2024

Replies: 5 comments 29 replies

GreatV
May 29, 2024
Maintainer

kishore-work-hard May 29, 2024
Author

kishore-work-hard
Jun 5, 2024
Author

GreatV
Jun 6, 2024
Maintainer

kishore-work-hard Jun 12, 2024
Author

kishore-work-hard Jun 12, 2024
Author

GreatV Jun 12, 2024
Maintainer

kishore-work-hard Jun 12, 2024
Author

kishore-work-hard Jun 12, 2024
Author

GreatV
Jun 13, 2024
Maintainer

GreatV Jun 13, 2024
Maintainer

kishore-work-hard Jun 13, 2024
Author

kishore-work-hard Jun 13, 2024
Author

GreatV Jun 13, 2024
Maintainer

kishore-work-hard Jun 13, 2024
Author

kishore-work-hard
Jun 14, 2024
Author