OCR问题
#51
by
william0014
- opened
模型在OCR提取文字过程中, 比如营业执照, 有较大概率识别营业执照号码不正确, 经常是中间少一位. 请问有什么办法能提高OCR精度?
Yes, I also met this. VLM has show great power and flexible ability，but they may make some very easy error. May SFT helps.