ZYM-PKU / UDiffText

UDiffText: A Unified Framework for High-quality Text Synthesis in Arbitrary Images via Character-aware Diffusion Models
MIT License
192 stars 16 forks source link

请教训练数据集问题 #3

Closed yang-chenyu104 closed 8 months ago

yang-chenyu104 commented 8 months ago

可以训练中文数据集吗

ZYM-PKU commented 8 months ago

理论上是可行的,但是需要具有单字级别标注的高质量数据集

yang-chenyu104 commented 8 months ago

那可以训练一个中文编码器,用文本检测器将文本区域检测,然后做一个中文文本编辑模型

---原始邮件--- 发件人: @.> 发送时间: 2023年12月20日(周三) 晚上9:05 收件人: @.>; 抄送: @.**@.>; 主题: Re: [ZYM-PKU/UDiffText] 请教训练数据集问题 (Issue #3)

理论上是可行的,但是需要具有单字级别标注的高质量数据集

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

yang-chenyu104 commented 8 months ago

谢谢大佬分享工作,启发很多

---原始邮件--- 发件人: @.> 发送时间: 2023年12月20日(周三) 晚上9:05 收件人: @.>; 抄送: @.**@.>; 主题: Re: [ZYM-PKU/UDiffText] 请教训练数据集问题 (Issue #3)

理论上是可行的,但是需要具有单字级别标注的高质量数据集

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

ZYM-PKU commented 8 months ago

这个想法很棒,我也会在接下来的工作中考虑类似的方法,感谢。

yang-chenyu104 commented 8 months ago

可以多多交流!!!

---原始邮件--- 发件人: @.> 发送时间: 2023年12月20日(周三) 晚上9:32 收件人: @.>; 抄送: @.**@.>; 主题: Re: [ZYM-PKU/UDiffText] 请教训练数据集问题 (Issue #3)

这个想法很棒,我也会在接下来的工作中考虑类似的方法,感谢。

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>