add ar models2 (#4604)

Co-authored-by: cuicheng01 <45199522+cuicheng01@users.noreply.github.com>

.precommit/check_imports.py

@@ -96,6 +96,7 @@ MOD_TO_DEP = {
     "vllm": "vllm",
     "xformers": "xformers",
     "yarl": "yarl",
+    "bidi": "python-bidi",
 }
 MOD_PATTERN = re.compile(
     rf"^(?:{'|'.join([re.escape(mod) for mod in MOD_TO_DEP])})(?=\.|$)"

docs/module_usage/tutorials/ocr_modules/text_recognition.en.md

@@ -310,6 +310,51 @@ el_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-mo
 <td>The Greek recognition model trained based on the PP-OCRv5 recognition model supports recognition of Greek, English, and numbers.</td>
 </tr>
 <tr>
+<td>arabic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/arabic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>81.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.6</td>
+<td>Ultra-lightweight Arabic character recognition model trained based on the PP-OCRv5 recognition model, supports Arabic letters and number recognition</td>
+</tr>
+<tr>
+<td>cyrillic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/cyrillic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>80.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.7</td>
+<td>Ultra-lightweight Cyrillic character recognition model trained based on the PP-OCRv5 recognition model, supports Cyrillic letters and number recognition</td>
+</tr>
+<tr>
+<td>devanagari_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/devanagari_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>84.96</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>Ultra-lightweight Devanagari script recognition model trained based on the PP-OCRv5 recognition model, supports Hindi, Sanskrit and other Devanagari letters, as well as number recognition</td>
+</tr>
+<tr>
+<td>te_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/te_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>87.65</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>Ultra-lightweight Telugu script recognition model trained based on the PP-OCRv5 recognition model, supports Telugu script and number recognition</td>
+</tr>
+<tr>
+<td>ta_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/ta_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>94.2</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>Ultra-lightweight Tamil script recognition model trained based on the PP-OCRv5 recognition model, supports Tamil script and number recognition</td>
+</tr>
+<tr>
 <td>korean_PP-OCRv3_mobile_rec</td>
 <td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/korean_PP-OCRv3_mobile_rec_infer.tar">Inference Model</a>/<a href="">Training Model</a></td>
 <td>60.21</td>

docs/module_usage/tutorials/ocr_modules/text_recognition.md

@@ -319,6 +319,56 @@ el_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model
 <td>基于PP-OCRv5识别模型训练得到的希腊语识别模型， 支持希腊语、英文和数字识别</td>
 </tr>
 <tr>
+<td>arabic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+arabic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>81.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.6</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量阿拉伯字母识别模型，支持阿拉伯字母、数字识别</td>
+</tr>
+<tr>
+<td>cyrillic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+cyrillic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>80.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.7</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量斯拉夫字母识别模型，支持斯拉夫字母、数字识别</td>
+</tr>
+<tr>
+<td>devanagari_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+devanagari_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>84.96</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量天城文识别模型，支持印地文、梵文等字母以及数字识别</td>
+</tr>
+<tr>
+<td>te_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+te_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>87.65</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量泰卢固文识别模型，支持泰卢固文、数字识别</td>
+</tr>
+<tr>
+<td>ta_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+ta_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>94.2</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量泰米尔文识别模型，支持泰米尔文、数字识别</td>
+</tr>
+<tr>
 <td>korean_PP-OCRv3_mobile_rec</td>
 <td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
 korean_PP-OCRv3_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/korean_PP-OCRv3_mobile_rec_pretrained.pdparams">训练模型</a></td>

docs/pipeline_usage/tutorials/ocr_pipelines/OCR.en.md

@@ -471,6 +471,51 @@ el_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-mo
 <td>The Greek recognition model trained based on the PP-OCRv5 recognition model supports recognition of Greek, English, and numbers.</td>
 </tr>
 <tr>
+<td>arabic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/arabic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>81.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.6</td>
+<td>Ultra-lightweight Arabic character recognition model trained based on the PP-OCRv5 recognition model, supports Arabic letters and number recognition</td>
+</tr>
+<tr>
+<td>cyrillic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/cyrillic_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>80.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.7</td>
+<td>Ultra-lightweight Cyrillic character recognition model trained based on the PP-OCRv5 recognition model, supports Cyrillic letters and number recognition</td>
+</tr>
+<tr>
+<td>devanagari_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/devanagari_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>84.96</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>Ultra-lightweight Devanagari script recognition model trained based on the PP-OCRv5 recognition model, supports Hindi, Sanskrit and other Devanagari letters, as well as number recognition</td>
+</tr>
+<tr>
+<td>te_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/te_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>87.65</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>Ultra-lightweight Telugu script recognition model trained based on the PP-OCRv5 recognition model, supports Telugu script and number recognition</td>
+</tr>
+<tr>
+<td>ta_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/ta_PP-OCRv5_mobile_rec_infer.tar">Inference Model</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">Pretrained Model</a></td>
+<td>94.2</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>Ultra-lightweight Tamil script recognition model trained based on the PP-OCRv5 recognition model, supports Tamil script and number recognition</td>
+</tr>
+<tr>
 <td>korean_PP-OCRv3_mobile_rec</td>
 <td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/korean_PP-OCRv3_mobile_rec_infer.tar">Inference Model</a>/<a href="">Training Model</a></td>
 <td>60.21</td>

docs/pipeline_usage/tutorials/ocr_pipelines/OCR.md

@@ -448,6 +448,56 @@ el_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model
 <td>基于PP-OCRv5识别模型训练得到的希腊语识别模型， 支持希腊语、英文和数字识别</td>
 </tr>
 <tr>
+<td>arabic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+arabic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>81.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.6</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量阿拉伯字母识别模型，支持阿拉伯字母、数字识别</td>
+</tr>
+<tr>
+<td>cyrillic_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+cyrillic_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>80.27</td>
+<td>-</td>
+<td>-</td>
+<td>7.7</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量斯拉夫字母识别模型，支持斯拉夫字母、数字识别</td>
+</tr>
+<tr>
+<td>devanagari_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+devanagari_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>84.96</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量天城文识别模型，支持印地文、梵文等字母以及数字识别</td>
+</tr>
+<tr>
+<td>te_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+te_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>87.65</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量泰卢固文识别模型，支持泰卢固文、数字识别</td>
+</tr>
+<tr>
+<td>ta_PP-OCRv5_mobile_rec</td>
+<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
+ta_PP-OCRv5_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams">训练模型</a></td>
+<td>94.2</td>
+<td>-</td>
+<td>-</td>
+<td>7.5</td>
+<td>基于PP-OCRv5识别模型训练得到的超轻量泰米尔文识别模型，支持泰米尔文、数字识别</td>
+</tr>
+<tr>
 <td>korean_PP-OCRv3_mobile_rec</td>
 <td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0.0/\
 korean_PP-OCRv3_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/korean_PP-OCRv3_mobile_rec_pretrained.pdparams">训练模型</a></td>

paddlex/configs/modules/text_recognition/arabic_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,39 @@
+Global:
+  model: arabic_PP-OCRv5_mobile_rec
+  mode: check_dataset # check_dataset/train/evaluate/predict
+  dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
+  device: gpu:0,1,2,3
+  output: "output"
+
+CheckDataset:
+  convert: 
+    enable: False
+    src_dataset_type: null
+  split: 
+    enable: False
+    train_percent: null
+    val_percent: null
+
+Train:
+  epochs_iters: 20
+  batch_size: 8
+  learning_rate: 0.001
+  pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams
+  resume_path: null
+  log_interval: 20
+  eval_interval: 1
+  save_interval: 1
+
+Evaluate:
+  weight_path: "output/best_accuracy/best_accuracy.pdparams"
+  log_interval: 1
+
+Export:
+  weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/arabic_PP-OCRv5_mobile_rec_pretrained.pdparams
+
+Predict:
+  batch_size: 1
+  model_dir: "output/best_accuracy/inference"
+  input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_011_arabic.png"
+  kernel_option:
+    run_mode: paddle

paddlex/configs/modules/text_recognition/cyrillic_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,39 @@
+Global:
+  model: cyrillic_PP-OCRv5_mobile_rec
+  mode: check_dataset # check_dataset/train/evaluate/predict
+  dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
+  device: gpu:0,1,2,3
+  output: "output"
+
+CheckDataset:
+  convert: 
+    enable: False
+    src_dataset_type: null
+  split: 
+    enable: False
+    train_percent: null
+    val_percent: null
+
+Train:
+  epochs_iters: 20
+  batch_size: 8
+  learning_rate: 0.001
+  pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams
+  resume_path: null
+  log_interval: 20
+  eval_interval: 1
+  save_interval: 1
+
+Evaluate:
+  weight_path: "output/best_accuracy/best_accuracy.pdparams"
+  log_interval: 1
+
+Export:
+  weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/cyrillic_PP-OCRv5_mobile_rec_pretrained.pdparams
+
+Predict:
+  batch_size: 1
+  model_dir: "output/best_accuracy/inference"
+  input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_011_cyrillic.png"
+  kernel_option:
+    run_mode: paddle

paddlex/configs/modules/text_recognition/devanagari_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,39 @@
+Global:
+  model: devanagari_PP-OCRv5_mobile_rec
+  mode: check_dataset # check_dataset/train/evaluate/predict
+  dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
+  device: gpu:0,1,2,3
+  output: "output"
+
+CheckDataset:
+  convert: 
+    enable: False
+    src_dataset_type: null
+  split: 
+    enable: False
+    train_percent: null
+    val_percent: null
+
+Train:
+  epochs_iters: 20
+  batch_size: 8
+  learning_rate: 0.001
+  pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams
+  resume_path: null
+  log_interval: 20
+  eval_interval: 1
+  save_interval: 1
+
+Evaluate:
+  weight_path: "output/best_accuracy/best_accuracy.pdparams"
+  log_interval: 1
+
+Export:
+  weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/devanagari_PP-OCRv5_mobile_rec_pretrained.pdparams
+
+Predict:
+  batch_size: 1
+  model_dir: "output/best_accuracy/inference"
+  input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_012_devanagari.png"
+  kernel_option:
+    run_mode: paddle

paddlex/configs/modules/text_recognition/ta_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,39 @@
+Global:
+  model: ta_PP-OCRv5_mobile_rec
+  mode: check_dataset # check_dataset/train/evaluate/predict
+  dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
+  device: gpu:0,1,2,3
+  output: "output"
+
+CheckDataset:
+  convert: 
+    enable: False
+    src_dataset_type: null
+  split: 
+    enable: False
+    train_percent: null
+    val_percent: null
+
+Train:
+  epochs_iters: 20
+  batch_size: 8
+  learning_rate: 0.001
+  pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams
+  resume_path: null
+  log_interval: 20
+  eval_interval: 1
+  save_interval: 1
+
+Evaluate:
+  weight_path: "output/best_accuracy/best_accuracy.pdparams"
+  log_interval: 1
+
+Export:
+  weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ta_PP-OCRv5_mobile_rec_pretrained.pdparams
+
+Predict:
+  batch_size: 1
+  model_dir: "output/best_accuracy/inference"
+  input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_008_ta.png"
+  kernel_option:
+    run_mode: paddle

paddlex/configs/modules/text_recognition/te_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,39 @@
+Global:
+  model: te_PP-OCRv5_mobile_rec
+  mode: check_dataset # check_dataset/train/evaluate/predict
+  dataset_dir: "/paddle/dataset/paddlex/ocr_rec/ocr_rec_dataset_examples"
+  device: gpu:0,1,2,3
+  output: "output"
+
+CheckDataset:
+  convert: 
+    enable: False
+    src_dataset_type: null
+  split: 
+    enable: False
+    train_percent: null
+    val_percent: null
+
+Train:
+  epochs_iters: 20
+  batch_size: 8
+  learning_rate: 0.001
+  pretrain_weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams
+  resume_path: null
+  log_interval: 20
+  eval_interval: 1
+  save_interval: 1
+
+Evaluate:
+  weight_path: "output/best_accuracy/best_accuracy.pdparams"
+  log_interval: 1
+
+Export:
+  weight_path: https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/te_PP-OCRv5_mobile_rec_pretrained.pdparams
+
+Predict:
+  batch_size: 1
+  model_dir: "output/best_accuracy/inference"
+  input: "https://paddle-model-ecology.bj.bcebos.com/paddlex/imgs/demo_image/general_ocr_rec_006_te.png"
+  kernel_option:
+    run_mode: paddle

paddlex/inference/models/text_recognition/predictor.py

@@ -15,6 +15,7 @@
 import numpy as np
 
 from ....modules.text_recognition.model_list import MODELS
+from ....utils.deps import class_requires_deps, is_dep_available
 from ....utils.fonts import (
     ARABIC_FONT,
     CYRILLIC_FONT,
@@ -35,7 +36,11 @@ from ..base import BasePredictor
 from .processors import CTCLabelDecode, OCRReisizeNormImg, ToBatch
 from .result import TextRecResult
 
+if is_dep_available("python-bidi"):
+    from bidi.algorithm import get_display
 
+
+@class_requires_deps("python-bidi")
 class TextRecPredictor(BasePredictor):
 
     entities = MODELS
@@ -104,6 +109,11 @@ class TextRecPredictor(BasePredictor):
             wh_ratio_list=wh_ratio_list,
             max_wh_ratio=max_wh_ratio,
         )
+        if self.model_name in (
+            "arabic_PP-OCRv3_mobile_rec",
+            "arabic_PP-OCRv5_mobile_rec",
+        ):
+            texts = [get_display(s) for s in texts]
         return {
             "input_path": batch_data.input_paths,
             "page_index": batch_data.page_indexes,
@@ -152,6 +162,7 @@ class TextRecPredictor(BasePredictor):
 
         if self.model_name in (
             "cyrillic_PP-OCRv3_mobile_rec",
+            "cyrillic_PP-OCRv5_mobile_rec",
             "eslav_PP-OCRv5_mobile_rec",
         ):
             return CYRILLIC_FONT
@@ -168,17 +179,23 @@ class TextRecPredictor(BasePredictor):
         if self.model_name == "el_PP-OCRv5_mobile_rec":
             return EL_FONT
 
-        if self.model_name == "arabic_PP-OCRv3_mobile_rec":
+        if self.model_name in (
+            "arabic_PP-OCRv3_mobile_rec",
+            "arabic_PP-OCRv5_mobile_rec",
+        ):
             return ARABIC_FONT
 
         if self.model_name == "ka_PP-OCRv3_mobile_rec":
             return KANNADA_FONT
 
-        if self.model_name == "te_PP-OCRv3_mobile_rec":
+        if self.model_name in ("te_PP-OCRv3_mobile_rec", "te_PP-OCRv5_mobile_rec"):
             return TELUGU_FONT
 
-        if self.model_name == "ta_PP-OCRv3_mobile_rec":
+        if self.model_name in ("ta_PP-OCRv3_mobile_rec", "ta_PP-OCRv5_mobile_rec"):
             return TAMIL_FONT
 
-        if self.model_name == "devanagari_PP-OCRv3_mobile_rec":
+        if self.model_name in (
+            "devanagari_PP-OCRv3_mobile_rec",
+            "devanagari_PP-OCRv5_mobile_rec",
+        ):
             return DEVANAGARI_FONT

paddlex/inference/utils/official_models.py

@@ -315,6 +315,11 @@ ALL_MODELS = [
     "th_PP-OCRv5_mobile_rec",
     "el_PP-OCRv5_mobile_rec",
     "en_PP-OCRv5_mobile_rec",
+    "arabic_PP-OCRv5_mobile_rec",
+    "te_PP-OCRv5_mobile_rec",
+    "ta_PP-OCRv5_mobile_rec",
+    "devanagari_PP-OCRv5_mobile_rec",
+    "cyrillic_PP-OCRv5_mobile_rec"
 ]
 
 
@@ -388,6 +393,11 @@ OCR_MODELS = [
     "te_PP-OCRv3_mobile_rec",
     "UniMERNet",
     "UVDoc",
+    "arabic_PP-OCRv5_mobile_rec",
+    "te_PP-OCRv5_mobile_rec",
+    "ta_PP-OCRv5_mobile_rec",
+    "devanagari_PP-OCRv5_mobile_rec",
+    "cyrillic_PP-OCRv5_mobile_rec"
 ]
 
 

paddlex/modules/text_recognition/model_list.py

@@ -39,4 +39,9 @@ MODELS = [
     "en_PP-OCRv5_mobile_rec",
     "el_PP-OCRv5_mobile_rec",
     "th_PP-OCRv5_mobile_rec",
+    "arabic_PP-OCRv5_mobile_rec",
+    "cyrillic_PP-OCRv5_mobile_rec",
+    "devanagari_PP-OCRv5_mobile_rec",
+    "ta_PP-OCRv5_mobile_rec",
+    "te_PP-OCRv5_mobile_rec"
 ]

paddlex/repo_apis/PaddleOCR_api/configs/arabic_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,141 @@
+Global:
+  model_name: arabic_PP-OCRv5_mobile_rec # To use static model for inference.
+  debug: false
+  use_gpu: true
+  epoch_num: 75
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/arabic_rec_ppocr_v5
+  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:
+  character_dict_path: ./ppocr/utils/dict/ppocrv5_arabic_dict.txt
+  max_text_length: &max_text_length 25
+  infer_mode: false
+  use_space_char: true
+  distributed: true
+  save_res_path: ./output/rec/predicts_arabic_ppocrv5.txt
+  d2s_train_image_shape: [3, 48, 320]
+
+
+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:
+  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: 0.00001
+      - NRTRHead:
+          nrtr_dim: 384
+          max_text_length: *max_text_length
+
+Loss:
+  name: MultiLoss
+  loss_config_list:
+    - CTCLoss:
+    - NRTRLoss:
+
+PostProcess:  
+  name: CTCLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  ignore_space: False
+
+Train:
+  dataset:
+    name: MultiScaleDataSet
+    ds_width: false
+    data_dir: ./train_data/
+    ext_op_transform_idx: 1
+    label_file_list:
+    - ./train_data/train_list.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:
+        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: &bs 128
+    fix_bs: false
+    divided_factor: [8, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: true
+    batch_size_per_card: *bs
+    drop_last: true
+    num_workers: 8
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data/
+    label_file_list:
+    - ./train_data/val_list.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: true
+    drop_last: false
+    batch_size_per_card: 128
+    num_workers: 4

paddlex/repo_apis/PaddleOCR_api/configs/cyrillic_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,141 @@
+Global:
+  model_name: cyrillic_PP-OCRv5_mobile_rec # To use static model for inference.
+  debug: false
+  use_gpu: true
+  epoch_num: 75
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/cyrillic_rec_ppocr_v5
+  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:
+  character_dict_path: ./ppocr/utils/dict/ppocrv5_cyrillic_dict.txt
+  max_text_length: &max_text_length 25
+  infer_mode: false
+  use_space_char: true
+  distributed: true
+  save_res_path: ./output/rec/predicts_cyrillic_ppocrv5.txt
+  d2s_train_image_shape: [3, 48, 320]
+
+
+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:
+  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: 0.00001
+      - NRTRHead:
+          nrtr_dim: 384
+          max_text_length: *max_text_length
+
+Loss:
+  name: MultiLoss
+  loss_config_list:
+    - CTCLoss:
+    - NRTRLoss:
+
+PostProcess:  
+  name: CTCLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  ignore_space: False
+
+Train:
+  dataset:
+    name: MultiScaleDataSet
+    ds_width: false
+    data_dir: ./train_data/
+    ext_op_transform_idx: 1
+    label_file_list:
+    - ./train_data/train_list.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:
+        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: &bs 128
+    fix_bs: false
+    divided_factor: [8, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: true
+    batch_size_per_card: *bs
+    drop_last: true
+    num_workers: 8
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data/
+    label_file_list:
+    - ./train_data/val_list.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: true
+    drop_last: false
+    batch_size_per_card: 128
+    num_workers: 4

paddlex/repo_apis/PaddleOCR_api/configs/devanagari_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,141 @@
+Global:
+  model_name: devanagari_PP-OCRv5_mobile_rec # To use static model for inference.
+  debug: false
+  use_gpu: true
+  epoch_num: 75
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/devanagari_rec_ppocr_v5
+  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:
+  character_dict_path: ./ppocr/utils/dict/ppocrv5_devanagari_dict.txt
+  max_text_length: &max_text_length 25
+  infer_mode: false
+  use_space_char: true
+  distributed: true
+  save_res_path: ./output/rec/predicts_devanagari_ppocrv5.txt
+  d2s_train_image_shape: [3, 48, 320]
+
+
+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:
+  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: 0.00001
+      - NRTRHead:
+          nrtr_dim: 384
+          max_text_length: *max_text_length
+
+Loss:
+  name: MultiLoss
+  loss_config_list:
+    - CTCLoss:
+    - NRTRLoss:
+
+PostProcess:  
+  name: CTCLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  ignore_space: False
+
+Train:
+  dataset:
+    name: MultiScaleDataSet
+    ds_width: false
+    data_dir: ./train_data/
+    ext_op_transform_idx: 1
+    label_file_list:
+    - ./train_data/train_list.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:
+        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: &bs 128
+    fix_bs: false
+    divided_factor: [8, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: true
+    batch_size_per_card: *bs
+    drop_last: true
+    num_workers: 8
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data/
+    label_file_list:
+    - ./train_data/val_list.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: true
+    drop_last: false
+    batch_size_per_card: 128
+    num_workers: 4

paddlex/repo_apis/PaddleOCR_api/configs/ta_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,141 @@
+Global:
+  model_name: ta_PP-OCRv5_mobile_rec # To use static model for inference.
+  debug: false
+  use_gpu: true
+  epoch_num: 75
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/ta_rec_ppocr_v5
+  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:
+  character_dict_path: ./ppocr/utils/dict/ppocrv5_ta_dict.txt
+  max_text_length: &max_text_length 25
+  infer_mode: false
+  use_space_char: true
+  distributed: true
+  save_res_path: ./output/rec/predicts_ta_ppocrv5.txt
+  d2s_train_image_shape: [3, 48, 320]
+
+
+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:
+  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: 0.00001
+      - NRTRHead:
+          nrtr_dim: 384
+          max_text_length: *max_text_length
+
+Loss:
+  name: MultiLoss
+  loss_config_list:
+    - CTCLoss:
+    - NRTRLoss:
+
+PostProcess:  
+  name: CTCLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  ignore_space: False
+
+Train:
+  dataset:
+    name: MultiScaleDataSet
+    ds_width: false
+    data_dir: ./train_data/
+    ext_op_transform_idx: 1
+    label_file_list:
+    - ./train_data/train_list.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:
+        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: &bs 128
+    fix_bs: false
+    divided_factor: [8, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: true
+    batch_size_per_card: *bs
+    drop_last: true
+    num_workers: 8
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data/
+    label_file_list:
+    - ./train_data/val_list.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: true
+    drop_last: false
+    batch_size_per_card: 128
+    num_workers: 4

paddlex/repo_apis/PaddleOCR_api/configs/te_PP-OCRv5_mobile_rec.yaml

@@ -0,0 +1,141 @@
+Global:
+  model_name: te_PP-OCRv5_mobile_rec # To use static model for inference.
+  debug: false
+  use_gpu: true
+  epoch_num: 75
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/te_rec_ppocr_v5
+  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:
+  character_dict_path: ./ppocr/utils/dict/ppocrv5_te_dict.txt
+  max_text_length: &max_text_length 25
+  infer_mode: false
+  use_space_char: true
+  distributed: true
+  save_res_path: ./output/rec/predicts_te_ppocrv5.txt
+  d2s_train_image_shape: [3, 48, 320]
+
+
+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:
+  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: 0.00001
+      - NRTRHead:
+          nrtr_dim: 384
+          max_text_length: *max_text_length
+
+Loss:
+  name: MultiLoss
+  loss_config_list:
+    - CTCLoss:
+    - NRTRLoss:
+
+PostProcess:  
+  name: CTCLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  ignore_space: False
+
+Train:
+  dataset:
+    name: MultiScaleDataSet
+    ds_width: false
+    data_dir: ./train_data/
+    ext_op_transform_idx: 1
+    label_file_list:
+    - ./train_data/train_list.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:
+        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: &bs 128
+    fix_bs: false
+    divided_factor: [8, 16] # w, h
+    is_training: True
+  loader:
+    shuffle: true
+    batch_size_per_card: *bs
+    drop_last: true
+    num_workers: 8
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data/
+    label_file_list:
+    - ./train_data/val_list.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: true
+    drop_last: false
+    batch_size_per_card: 128
+    num_workers: 4

paddlex/repo_apis/PaddleOCR_api/text_rec/register.py

@@ -268,3 +268,48 @@ register_model_info(
         "supported_apis": ["train", "evaluate", "predict", "export"],
     }
 )
+
+register_model_info(
+    {
+        "model_name": "arabic_PP-OCRv5_mobile_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "arabic_PP-OCRv5_mobile_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)
+
+register_model_info(
+    {
+        "model_name": "te_PP-OCRv5_mobile_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "te_PP-OCRv5_mobile_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)
+
+register_model_info(
+    {
+        "model_name": "ta_PP-OCRv5_mobile_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "ta_PP-OCRv5_mobile_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)
+
+register_model_info(
+    {
+        "model_name": "devanagari_PP-OCRv5_mobile_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "devanagari_PP-OCRv5_mobile_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)
+
+register_model_info(
+    {
+        "model_name": "cyrillic_PP-OCRv5_mobile_rec",
+        "suite": "TextRec",
+        "config_path": osp.join(PDX_CONFIG_DIR, "cyrillic_PP-OCRv5_mobile_rec.yaml"),
+        "supported_apis": ["train", "evaluate", "predict", "export"],
+    }
+)

setup.py

@@ -183,6 +183,7 @@ EXTRAS = {
             "shapely",
             "tiktoken",
             "tokenizers",
+            "python-bidi",
         ],
         "speech": [
             "ftfy",