Ucas-HaoranWei / GOT-OCR2.0

Official code implementation of General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model
6.12k stars 526 forks source link

Several format failure #60

Open AdamPlatin123 opened 2 months ago

AdamPlatin123 commented 2 months ago

捉个虫 这种识别似乎对于带有大量横线的文档(比如填空题)有点bug,经常识别成表格。 image 识别结果如下(多种模式尝试结果相同) image

Ucas-HaoranWei commented 2 months ago

这种数据需要你来微调一下,我这边训练数据比较粗糙

AdamPlatin123 commented 2 months ago

感谢回复,我比较好奇的是您对于模型采用了什么训练策略?我目前还在研究生学习阶段,涉及LLM的训练和微调,但对于图像模型的训练尚无经验。请问您能否在方便时提供相应训练平台等的链接等等信息。感激不尽。


发件人: WeiHaoran @.> 发送时间: 2024年9月20日 11:18 收件人: Ucas-HaoranWei/GOT-OCR2.0 @.> 抄送: AdamPlatin123 @.>; Author @.> 主题: Re: [Ucas-HaoranWei/GOT-OCR2.0] Several format failure (Issue #60)

这种数据需要你来微调一下,我这边训练数据比较粗糙

― Reply to this email directly, view it on GitHubhttps://github.com/Ucas-HaoranWei/GOT-OCR2.0/issues/60#issuecomment-2362678792, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BE574LFKJSHG5L6VHYQMPSLZXOHZVAVCNFSM6AAAAABOQRDMQ2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGNRSGY3TQNZZGI. You are receiving this because you authored the thread.Message ID: @.***>