Skip to content

PPStructureV3在Nvidia显卡下表格识别上下颠倒 #16417

@adamzhg

Description

@adamzhg

🔎 Search before asking

  • I have searched the PaddleOCR Docs and found no similar bug report.
  • I have searched the PaddleOCR Issues and found no similar bug report.
  • I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

使用PPStructureV3在英伟达GPU卡上推断一个表格,表格内容上下颠倒,在海光DCU上没有这个问题。
1、pipline的yaml文件中如果将table中ocr的配置项use_textline_orientation设为False,则表格识别结果文字识别大面积错误。打开这个选项文字识别不再出错,但是识别内容上下颠倒
2、在海光DCU上没有这个问题
3、paddle3.x试了几个组合(paddleocr、paddlex),都有这个问题

🏃‍♂️ Environment (运行环境)

OS:ubuntu22.04
paddleocr 3.2.0
paddlepaddle-gpu 3.1.0
paddlex 3.2.1
cuda:12.9

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

出错的图片:

Image

识别结果:
上面的【计量单位...】和下面的【接收单位经办人...】上下颠倒了

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions