Replies: 5 comments
-
可以增加相关数据进行微调试试,可能训练数据里数据比较少,效果比较差一些 |
Beta Was this translation helpful? Give feedback.
-
这个能在本地部署嘛 |
Beta Was this translation helpful? Give feedback.
-
可以的
…---原始邮件---
发件人: ***@***.***>
发送时间: 2025年4月14日(周一) 下午2:56
收件人: ***@***.***>;
抄送: ***@***.******@***.***>;
主题: Re: [PaddlePaddle/PaddleOCR] 文本识别错误 (Issue #15007)
可以增加相关数据进行微调试试,可能训练数据里数据比较少,效果比较差一些
这个能在本地部署嘛
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
bigBoLang left a comment (PaddlePaddle/PaddleOCR#15007)
可以增加相关数据进行微调试试,可能训练数据里数据比较少,效果比较差一些
这个能在本地部署嘛
—
Reply to this email directly, view it on GitHub, or unsubscribe.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
第一次接触,有教程可以学一下的嘛 |
Beta Was this translation helpful? Give feedback.
-
根据你提供的描述(如将“丐”识别为“正”),以及 PaddleOCR 相关 Issue 的参考(比如 字符表里面有的字,就是识别不出来 #14065),这种现象属于 PaddleOCR 识别模型的泛化能力或字典覆盖/训练数据不足导致的常见问题。具体分析和可能的解决方案如下: 原因分析
参考相关社区问题
解决思路与建议
总结综上,你遇到的“丐”被识别为“正”的问题,是 PaddleOCR 目前的已知短板,主要由训练样本不足和高频字干扰导致。如需高质量定制识别,建议自制数据集进行模型微调。更多细节可参考该 相关社区问答和讨论。 Response generated by 🤖 feifei-bot | gpt-4.1 |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
🔎 Search before asking
🐛 Bug (问题描述)
C:\Users\Administrator\xiayi\imgs '丐'识别成'正'
🏃♂️ Environment (运行环境)
win11
python 3.12
🌰 Minimal Reproducible Example (最小可复现问题的Demo)
C:\Users\Administrator\xiayi\imgs
Beta Was this translation helpful? Give feedback.
All reactions