latest/version3.x/pipeline_usage/PP-DocTranslation

[giscus[bot]]

latest/version3.x/pipeline_usage/PP-DocTranslation

Awesome multilingual OCR toolkits based on PaddlePaddle (practical ultra lightweight OCR system, support 80+ languages recognition, provide data annotation and synthesis tools, support training and deployment among server, mobile, embedded and IoT devices)

https://www.paddleocr.ai/latest/version3.x/pipeline_usage/PP-DocTranslation.html

[quanysq]

运行 2.2 Python脚本方式集成 中的代码，报错：
File "C:\Users\51398.conda\envs\Python312\Lib\site-packages\paddlex\inference\pipelines\layout_parsing\pipeline_v2.py", line 1186, in predict
self.seal_recognition_pipeline(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: '_LayoutParsingPipelineV2' object has no attribute 'seal_recognition_pipeline'. Did you mean: 'table_recognition_pipeline'?

[liuhongen1234567]

您好，可以将 use_seal_recognition=False试试, 最新版印章识别存在问题，我们会在下一版修复

    visual_predict_res = pipeline.visual_predict(
        input_path,
        use_doc_orientation_classify=False,
        use_doc_unwarping=False,
        use_common_ocr=True,
        use_seal_recognition=False,
        use_table_recognition=True,
    )

[refrainD]

运行2.2 Python脚本方式集成中的代码，当input_file为.md而不是pdf时，会报错:

Split the original text into 47 blocks
Starting translation...
Translating block 13/47...
Connection error.
Traceback (most recent call last):
File "/opt/paddle/PP-DocTransaltion/./api.py", line 52, in
tgt_md_info_list = translation_pipeline.translate(
File "/usr/local/lib/python3.10/dist-packages/paddleocr/_pipelines/pp_doctranslation.py", line 295, in translate
return list(
File "/usr/local/lib/python3.10/dist-packages/paddlex/inference/pipelines/pp_doctranslation/pipeline.py", line 478, in translate
target_language_texts = self.chunk_translate(
File "/usr/local/lib/python3.10/dist-packages/paddlex/inference/pipelines/pp_doctranslation/pipeline.py", line 327, in chunk_translate
translation_results.append(translate_func(chunk.strip()))
File "/usr/local/lib/python3.10/dist-packages/paddlex/inference/pipelines/pp_doctranslation/pipeline.py", line 452, in translate_func
raise Exception("The call to the large model failed.")
Exception: The call to the large model failed.

备注: llm控制台上可以看到system提示词的输入，却没有markdown内容输入。

我看到脚本中pipeline.visual_predict和pipeline.load_from_markdown这两个方法构建的ori_md_info_list数据格式不相同，经过visual_predict后构建的markdown列表可以用于translate，但load_from_markdown却会报错，是否是这个原因？

如果我只需要使用translation产线的translate翻译功能，是否一定需要先使用PP-DocTranslation产线的visual_predict获取视觉预测结果？ visual_predict是否像StructureV3产线一样提供了lang语言参数?
能否自定义python脚本先调用StructureV3产线的版面解析在构建参数执行PP-DocTranslation产线的翻译，从而为不同语言指定文本解析的ocr模型.

[refrainD]

• • 修改模型链接参数已经解决上述问题了，请问PP-DocTranslation产线的 visual_predict方法 有没有提供lang语言参数，源码中没有找到。