firefox12
V2EX  ›  问与答

pdf ocr 用什么应用或者 ai?

  •  
  •   firefox12 · Sep 27 · 1217 views
    This topic created in 43 days ago, the information mentioned may be changed or developed.

    我手里有一些学习资料,就是扫描后考卷,我想把它的内容 ocr 成为文档,但是我发现我用 umi 或者在线的 lovepdf 这种 出来的结果 ,格式错字都很多。格式基本上不能看。 我就想问,ai 这么强了 有能用的 pdf ocr 吗?

    可付费解决这个问题。

    https://demo1.k8seasy.com/sample.pdf

    这里有个简单的 pdf 内容基本就是这样的,每一课都有答案和解答。 希望能 ocr 成 可以直接使用的内容,现在的题目里面的文字都有大量缺失。根本没办法用

    12 replies  •  2026-11-04 18:34:04 +08:00
    k74qr
    k74qr
    Sep 28❤️1
    1
    试试 paddleocr 和 mineru
    datocp
    datocp
    Sep 28 via Android
    2
    估计要花钱,手机端即便 google 的 app 也会出现因为角度问题将字识别错的问题。有个叫 text fairy 稍微好上那么一点点,会扫描后对齐,依然会出错。

    只有国产的那些据说是云扫描的,确实试用的时候那些会出错的位置,它都不会出错,因为要钱,后来听书干脆下 pdf 文件。
    firefox12
    firefox12
    OP
    Sep 28
    3
    @k74qr paddleocr 和 mineru 都试了 的确比其他的效果要好很多,虽然还不算完美 但是 缺字 漏字的情况少了 80%是有的
    GPLer
    GPLer
    Sep 28 via Android
    4
    量不大的话推荐用多模态 AI 来识别,生成的格式会更好,还有一定的纠错能力。
    leir
    leir
    Sep 28
    5
    试了一下使用 paddleocr 只能搞定 95%吧,有预算还是得上 AI
    firefox12
    firefox12
    OP
    Sep 28
    6
    ai 怎么弄?

    用 paddleocr 转了一下 90% 基本都是对的, 因为字本来也很清晰,但是排版还是游戏和原版的差别。算了 剩下的写了一个 py 校验了一下,手动改了 2 小时,第一版算是弄出来了。
    moskize
    moskize
    Oct 21
    7
    我刚用 pdfcraft.ai 转了一下。

    ```
    # 进阶练 01 动词时态一(模拟真题综合汇编)

    学校: 姓名: 班级: 考号:一般现在时、一般过去时、一般将来时、现在进行时、过去进行时

    ### 一、单项选择

    ## 一般现在时

    1. ( 2023·上海松江·统考二模) Jane a DIY card to her Mum on Mother's Day every year.

    A. sends B. is sending C. has sent D. sent

    2. ( 2023·上海徐汇·统考一模) A couple of young people cycling in the countryside once a month every year.

    A. go B. went C. will go D. has gone

    3. ( 2023·上海·九年级专题练习) your brother always breakfast at seven every morning?

    A. Does, eat B. Does, eats C. Do, eat D. Do, eats

    4. ( 2022 秋·上海浦东新·九年级上海市实验学校校考期末) If you take a fish out of water, itA. will be dead B. would die C. dies D. died

    5. ( 2022·上海浦东新·统考模拟预测) Some adults to the hospital to have a physical checkup once a year.

    A. go B. have gone C. are going D. went

    6. ( 2022·上海普陀·统考二模) Wendy two cups of coffee in the garden every morning.

    A. drink B. is drinking C. drinks D. drank

    7. ( 2022·上海闵行·统考二模) My father the underground to work on weekdays. He is a green consumer.

    A. take B. takes C. took D. has taken

    8. ( 2022·上海嘉定·统考二模) As a local journalist, John big events of the city every day.

    A. will report B. reports C. was reporting D. reported

    9. ( 2022·上海徐汇·统考二模) As a green consumer, Tim usually around the city by public transport.

    A. travels B. travelled C. will travel D. is travelling

    10. ( 2022·上海崇明·统考一模) Lynn a bike to work on weekdays. She thinks it can help protect the environment.

    A. ride B. rides C. rode D. has ridden

    11. (2022·上海浦东·统考一模) I've heard that Tom a visit to his disabled father twice a week.

    A. pays B. paid C. will pay D. has paid

    12. (2022·上海·模拟预测) Kitty's dog rarely barks at people,

    A. does it B. doesn't it C. has it D. hasn't it

    13. (2022·上海·模拟预测) If you nervous before exams, you can take a deep breath and calm yourself down.

    A. is getting B. will get C. get D. got

    14. (2022·上海·徐汇·九年级上海市第四中学校考期末) My grandmother to the hospital to have a physical checkup twice a year.

    A. goes B. has gone C. is going D. went

    15. (2021·上海松江·统考二模) Tim likes watching films. He to the cinema with his girlfriend once a week.

    A. goes B. is going C. has gone D. will go

    16. (2021·上海青浦·统考二模) Please tell John to call me back as soon as he back to the office.

    A. come B. comes C. will come D. came

    17. (2021·上海徐汇·统考一模) Dan often in his car to fish with his friends on Sunday afternoon.

    A. sets off B. will set off C. has set off D. is setting off

    18. (2020·上海青浦·统考一模) People from Thailand never greet each other by shaking hands,A. don't they B. do they C. won't they D. will they

    19. (2020·上海·统考中考真题) Every year thousands of tourists the mountain area to relax themselves.

    A. visited B. were visiting C. visit D. have visited

    20. (2020·上海杨浦·统考二模) It took me an hour to get to school this morning. Most days it so long.

    A. not take B. doesn't take C. didn't take D. isn't taking

    21. (2020·上海虹口·统考二模) John promises that he will give his cousin a call as soon as he in Europe.

    A. arrived B. had arrived C. will arrive D. arrives

    22. (2020·上海杨浦·统考一模) We usually go far a walk after supper if there a good program on TV.

    A. isn't B. wasn't C. won't be D. wasn't going to be

    23. (2020·上海嘉定·统考一模) The student's won't start the experiment until the teacher them how to do it.

    A. will show B. show C. showed D. shows

    24. (2020·上海普陀·统考一模) Terry gets up at 6:30 every morning and then he a bath.

    A. has B. have C. had D. is having

    25. (2019·上海长宁·统考二模) Please tell Daisy to call me when she ______ back to Shanghai tomorrow afternoon.

    A. comes B. came C. will come D. have come

    ## 一般过去时

    26. (2023·上海杨浦·统考三模) —Have you seen Mike these days? —He ______ here the other day.

    A. has been B. had been C. was D. will be

    27. (2023·上海奉贤·统考二模) John, as well as his parents, ______ proud to support his community during last year's pandemic.

    A. is B. are C. was D. were

    28. (2023·上海徐汇·统考二模) Merle first ______ the world record for swimming 10 kilometers in 2019.

    A. sets B. set C. was setting D. has set

    29. (2023·上海普陀·统考二模) The boy ______ and showed us a row of white teeth.

    A. laugh B. laughed C. laughs D. is laughing

    30. (2023·上海·模拟预测) After the interview, Pansy ______ a report for the school newspaper last week.

    A. write B. writes C. wrote D. was writing

    31. (2023·上海·一模) Bena ______ detective story written by Agatha Christie last night.

    A. read B. is reading C. reads D. had read

    32. (2023·上海青浦·统考一模) Last weekend, Leo ______ himself by accident when he was cooking lunch.

    A. burns B. burnt C. was burning D. had burnt

    33. (2023·上海·一模) The three astronauts finally ______ after 183 days in space on 4, December 2022.

    A. returned B. are returning C. have returned D. will return

    34. (2023·上海长宁·统考一模) Tina hardly knew anything about yesterday's school sports meeting, ______ she?

    A. does B. doesn't C. did D. didn't

    35. (2023·上海杨浦·统考二模) My father ______ me a funny joke and I can't stop laughing every time I think of it.

    A. told B. tells C. will tell D. is telling
    ```
    firefox12
    firefox12
    OP
    Oct 26
    8
    @moskize 我最后用了 mineru 加手写,paddleocr 效果好像更好些,但是 它的内容 转成 txt 格式好像没 mineru 好
    moskize
    moskize
    Oct 27
    9
    @firefox12 手写是你手动校验吗?😂 你说 txt 格式没 mineru 好是啥意思?
    firefox12
    firefox12
    OP
    Oct 28
    10
    @moskize 因为我最终的目的是 写了一个本地的 app ,用户可以直接使用 ipad 打开了 做练习,得分,所以最终还是需要

    36. 35. (2023·上海杨浦·统考二模) My father ______ me a funny joke and I can't stop laughing every time I think of it.

    A. told B. tells C. will tell D. is telling
    这样 提示 A B C D 用户点击一个 判断对错。 提示结果。
    ocr 出来其实还是有不少错的, 比如 35. 它会丢失, 那么我的答案就没办法 一一匹配,

    所以 我让 ai 写个脚本 扫描 txt 按照一定的格式,1 2 3 4 如果 断开 乱了, 就提示 我再手动改改, 我可以让 ai 做,但是做出来 可能会把题目都变掉,所以 基本上还是用脚本跑比较准确。 大概 改了 70-80 个小错吧, 错误来自 缺少信息 和 少行。

    txt 格式没 mineru 好,就是 它给了一个复杂的 json 我用脚本都没办法解析,它的格式不规则,很难直接处理,还是 mineru 直接给 txt 好。
    moskize
    moskize
    5 days ago❤️1
    11
    @firefox12 你用 https://github.com/oomol-lab/pdf-craft 这个,就是我说的 pdfcraft.ai 的开源。

    你用程序生成 EPUB 格式,他们本质就是 ZIP 封装的 XML 文件。pdf-craft 翻译 XML 文件的时候会保留 XML 结构本身。就是说 <div>My <span>father</span> told me...</div> 会被翻译成 <div>我的<span>父亲</span>给我说了...</div>。不要用 markdown ,会丢掉这些细节。
    firefox12
    firefox12
    OP
    5 days ago
    12
    好的,下一次我再试试, 比较麻烦的就是 <div> <span> 这种很复杂的结构 我不知道如何用 python 统一处理,每道题的结构不一样 处理起来就很难
    About   ·   Help   ·   Advertise   ·   Blog   ·   API   ·   FAQ   ·   Privacy   ·   Solana   ·   948 Online   Highest 6679   ·     Select Language
    创意工作者们的社区
    World is powered by solitude
    VERSION: 3.9.8.5 · 35ms · UTC 20:03 · PVG 04:03 · LAX 12:03 · JFK 15:03
    ♥ Do have faith in what you're doing.