Fine-tuned Model not working, model mismatch error

[itzananthakrishnan]

Created a fine-tuned model using PaddlePaddle 3.0.0 GPU and PaddleOCR 3.0.1 with en_PPOCRv4_mobile_rec. and after that exported the model adn when trying to infer the model with :

from paddleocr import PaddleOCR
import os
import re

Input and Output folders

images_folder = "./images"
output_folder = "./test_out"
os.makedirs(output_folder, exist_ok=True)

Initialize PaddleOCR

ocr = PaddleOCR(
lang='en',
use_doc_orientation_classify=False,
use_textline_orientation=False,
device="gpu",
text_detection_model_name="PP-OCRv4_mobile_det",
text_recognition_model_dir = '/home/paddle_OCR/inference/en_PP-OCRv4_mobile_rec_infer'

)......
...
...

the error says :
UserWarning: No ccache found. Please be aware that recompiling all source files may be required. You can download and install ccache from: https://github.com/ccache/ccache/blob/master/doc/INSTALL.md
warnings.warn(warning_message)
Creating model: ('UVDoc', None)
Using official model (UVDoc), the model files will be automatically downloaded and saved in /home/trois/.paddlex/official_models.
Fetching 6 files: 100%|█████████████████████████████████████████████████████████████████| 6/6 [00:00<00:00, 47482.69it/s]
Creating model: ('PP-OCRv4_mobile_det', None)
Using official model (PP-OCRv4_mobile_det), the model files will be automatically downloaded and saved in /home/.paddlex/official_models.
Fetching 6 files: 100%|██████████████████████████████████████████████████████████████████| 6/6 [00:00<00:00, 7478.70it/s]
Creating model: ('PP-OCRv5_server_rec', '/home/paddle_OCR/inference/en_PP-OCRv4_mobile_rec_infer')
Traceback (most recent call last):

Even though I trained on OCR4_mobile_rec, it’s creating PP-OCRv5_server_rec.

And also, finally with this error:

model_name == config["Global"]["model_name"]
AssertionError: Model name mismatch, please input the correct model dir.

Can someone provide the script for infer?

[liuhongen1234567]

Hello, PP-OCRv5_server_rec is the default model for paddleocr 3.0. If you want to specify a different model, you need to set the text_recognition_model_name. The example code is as follows:

ocr = PaddleOCR(
lang='en',
use_doc_orientation_classify=False,
use_textline_orientation=False,
device="gpu",
text_detection_model_name="PP-OCRv4_mobile_det",
text_recognition_model_name="en_PP-OCRv4_mobile_rec",
text_recognition_model_dir = '/home/paddle_OCR/inference/en_PP-OCRv4_mobile_rec_infer'
)..

[itzananthakrishnan]

thanks, also Is there any way to turn off detection like previous versions? also the finetuned model is performing worse than the pretrained model? any specific reason / method is hould follow ? . because i was getting validation accuracy of 92% when training.

this is the en-ppocrv4_mobile_rec.yml i used for training :

Global:
model_name: en_PP-OCRv4_mobile_rec # To use static model for inference.
debug: false
use_gpu: true
epoch_num: 500
log_smooth_window: 20
print_batch_step: 10
save_model_dir: ./output/rec_ppocr_v4_custom
save_epoch_step: 10
eval_batch_step:

• 0
• 1000
  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: /home/paddle_OCR/data/en_dict.txt
  max_text_length: 25
  infer_mode: false
  use_space_char: true
  distributed: true
  save_res_path: ./output/rec/predicts_ppocrv3.txt
  Optimizer:
  name: Adam
  beta1: 0.9
  beta2: 0.999
  lr:
  name: Cosine
  learning_rate: 0.0005
  warmup_epoch: 5
  regularizer:
  name: L2
  factor: 3.0e-05
  Architecture:
  model_type: rec
  algorithm: SVTR_LCNet
  Transform: null
  Backbone:
  name: PPLCNetV3
  scale: 0.95
  Head:
  name: MultiHead
  head_list:
  • CTCHead:
    Neck:
    name: svtr
    dims: 120
    depth: 2
    hidden_dims: 120
    kernel_size:
    - 1
    - 3
    use_guide: true
    Head:
    fc_decay: 1.0e-05
  • NRTRHead:
    nrtr_dim: 384
    max_text_length: 25
    Loss:
    name: MultiLoss
    loss_config_list:
• CTCLoss: null
• NRTRLoss: null
  PostProcess:
  name: CTCLabelDecode
  Metric:
  name: RecMetric
  main_indicator: acc
  ignore_space: false
  Train:
  dataset:
  name: MultiScaleDataSet
  ds_width: false
  data_dir: /home/paddle_OCR/data/training_17-06-25_test/train
  ext_op_transform_idx: 1
  label_file_list:
  • /home/paddle_OCR/data/training_17-06-25_test/train/train.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: 25
  • RecAug: null
  • MultiLabelEncode:
    gtc_encode: NRTRLabelEncode
  • KeepKeys:
    keep_keys:
    • image
    • label_ctc
    • label_gtc
    • length
    • valid_ratio
      sampler:
      name: MultiScaleSampler
      scales:
  • • 320
    • 32
  • • 320
    • 48
  • • 320
    • 64
      first_bs: 96
      fix_bs: false
      divided_factor:
  • 8
  • 16
    is_training: true
    loader:
    shuffle: true
    batch_size_per_card: 96
    drop_last: true
    num_workers: 8
    Eval:
    dataset:
    name: SimpleDataSet
    data_dir: /home/paddle_OCR/data/training_17-06-25_test/val
    label_file_list:
  • /home/paddle_OCR/data/training_17-06-25_test/val/val.txt
    transforms:
  • DecodeImage:
    img_mode: BGR
    channel_first: false
  • MultiLabelEncode:
    gtc_encode: NRTRLabelEncode
  • RecResizeImg:
    image_shape:
    • 3
    • 48
    • 320
  • KeepKeys:
    keep_keys:
    • image
    • label_ctc
    • label_gtc
    • length
    • valid_ratio
      loader:
      shuffle: false
      drop_last: false
      batch_size_per_card: 128
      num_workers: 4
      profiler_options: null

The training procedure I followed:

Created text file with the image path and word using gen_label.py
Model training is completed, and I exported it and am using that now.