Extract Text from PDFs & Images for LLMs Using Python

แชร์
ฝัง
  • เผยแพร่เมื่อ 9 พ.ย. 2024

ความคิดเห็น • 71

  • @techwithzoum
    @techwithzoum  ปีที่แล้ว

    Please subscribe and like the video to help me keep motivated to make awesome videos like this one. :)

  • @eloghosaikponmwoba4452
    @eloghosaikponmwoba4452 4 หลายเดือนก่อน +1

    Thank you for this video. I needed this information at this very moment.

    • @techwithzoum
      @techwithzoum  4 หลายเดือนก่อน

      I am really happy it helped!

  • @AbdulAhad-Family
    @AbdulAhad-Family ปีที่แล้ว +1

    Fantastic tutorial, so much simplified ...great job

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      You are welcome, AbdoulAhad-Family!

  • @SMCGPRA
    @SMCGPRA 3 หลายเดือนก่อน +2

    The voice is very clear and crisp for non-english viewers to understand. Content is very excellent and explained exquisitely. Could you let us know if tables in the pdf or word doc, using RAG and prompt can we able to join tables, filter data from tables and other operations in Gen AI?

    • @techwithzoum
      @techwithzoum  3 หลายเดือนก่อน +1

      Thank you for your response!
      My course on Multimodal RAG below would be a good option to explore and might be what you are looking for to answer your questions.
      th-cam.com/video/D5iKsvK7cXg/w-d-xo.htmlsi=scyeB48fyvrn9nsU

    • @SMCGPRA
      @SMCGPRA 3 หลายเดือนก่อน +1

      @@techwithzoum thanks good to explore

    • @techwithzoum
      @techwithzoum  3 หลายเดือนก่อน

      @@SMCGPRA you are welcome!

  • @kenchang3456
    @kenchang3456 9 หลายเดือนก่อน

    Excellent tutorial, thank you. Your video was referenced on AI Jason's channel and I am glad he did.

  • @AIJasonZ
    @AIJasonZ ปีที่แล้ว +1

    This is awesome, great work!

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      Thank you, AI JASON!

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      Please, feel free to share with your audience for better outreach :)

  • @hasnainzafar7304
    @hasnainzafar7304 6 หลายเดือนก่อน +1

    Great work. Absolute genius!

  • @carlosedulira
    @carlosedulira 10 หลายเดือนก่อน +1

    So much thanks :) Great work!

  • @thibauteka4046
    @thibauteka4046 3 หลายเดือนก่อน +1

    Thanks so much. can we have an example for a data extration from a table on an image ?

    • @techwithzoum
      @techwithzoum  3 หลายเดือนก่อน

      Thanks, @thibauteka4046, and that sounds like a great idea!

  • @damienromy7460
    @damienromy7460 2 หลายเดือนก่อน +1

    Hello, what is your version of langchain i think mine doesnt work

    • @techwithzoum
      @techwithzoum  2 หลายเดือนก่อน

      Can you please share the error you are getting?

  • @amoghahalhalli8198
    @amoghahalhalli8198 หลายเดือนก่อน

    How is LlamaIndex PDF to Text Conversion as compared to any of these methods?

  • @rashmikasaha2874
    @rashmikasaha2874 7 หลายเดือนก่อน

    Hey, im getting an error
    ~\anaconda3\lib\genericpath.py in isfile(path)
    28 """Test whether a path is a regular file"""
    29 try:
    ---> 30 st = os.stat(path)
    31 except (OSError, ValueError):
    32 return False
    TypeError: stat: path should be string, bytes, os.PathLike or integer, not JpegImageFile
    can you help me with a fix?

  • @Abhi_interiors
    @Abhi_interiors ปีที่แล้ว +1

    which among these has the best accuracy?

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      The Unstructured library seems to do a great job since it has the features to handle any type of file without much difficulty.

  • @Yanikikudon
    @Yanikikudon ปีที่แล้ว

    Great Work, thank you ! 😀

  • @SuiGio
    @SuiGio ปีที่แล้ว

    Important question: If a pdf does have a picture in it, when converting to picture, firstly, is that picture added as text or is it skipped?
    Secondly, is there a way to *know* that the extracted text is coming from an image within the pdf? Some sort of metadata at least to get that info?
    Thanks for the video, nice content with overall breadth, wish you could answer my question.

    • @piyushchhawachharia_0068
      @piyushchhawachharia_0068 10 หลายเดือนก่อน

      Hi , Could you solve this issue?

    • @SuiGio
      @SuiGio 9 หลายเดือนก่อน

      @@piyushchhawachharia_0068 Hi, not really. I realised the solution that's usually done with the python libraries are that they pitcture-ize the entire pdf page. That means that pictures within pdfs, are read as characters but obviously semantics from visual representations are not transcribed. Thankfully the order of text appearing is good enough, haven't seen anything terribly wrong in the order of the text extracted. How do you approach this yourself?

    • @abhijeetsinha1803
      @abhijeetsinha1803 7 หลายเดือนก่อน

      Please suggest , i have a multiple pdfs and some of the pdfs have the images. i don't want to work and apply multimodel ai seperaterately for the pdfs which having images , i want to proceed like : load the documents -> split into multiple chunks -> embedding each chunks and stored into vector stores .. please suggest the process

  • @susmitsekhar5100
    @susmitsekhar5100 ปีที่แล้ว +1

    Great work. can we extract information from charts like histogram/barplot ?

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว +1

      Susmitsekhar, thank you!
      Yes, we can extract any textual information from histogram/barplot. Give it a try and I am happy to know more about your findings.

    • @susmitsekhar5100
      @susmitsekhar5100 ปีที่แล้ว +1

      @@techwithzoum Thanks for your quick response.Unfortunately it's not working for Histogram/barplot kind of things. Do you have any solutions.

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว +1

      @@susmitsekhar5100 I haven't explored any type of case like this, so can't tell right now.

    • @susmitsekhar5100
      @susmitsekhar5100 ปีที่แล้ว +1

      @@techwithzoum thanks for your response. kindly let us know if you come across any similar kind of things.

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว +1

      @@susmitsekhar5100 sure!
      Don't forget to subscribe and share the link of the channel with your friends!

  • @ThimaliFernando
    @ThimaliFernando 2 หลายเดือนก่อน +1

    Are these working for scanned pdfs?

    • @techwithzoum
      @techwithzoum  2 หลายเดือนก่อน

      Yes, they do

  • @vivekpatel2736
    @vivekpatel2736 5 หลายเดือนก่อน +1

    @zoumdatascience Can we give the image output with text based on the questions?

    • @techwithzoum
      @techwithzoum  5 หลายเดือนก่อน

      Yes, you can do that.

    • @vivekpatel2736
      @vivekpatel2736 5 หลายเดือนก่อน

      ​@@techwithzoumhow can i do that do you have any reference ?

    • @vivekpatel2736
      @vivekpatel2736 5 หลายเดือนก่อน

      Actually i need to make chat with pdf and want to get answer in text with image

  • @jonyswe580
    @jonyswe580 ปีที่แล้ว

    Good one!

  • @QorQar
    @QorQar ปีที่แล้ว

    مااسم صفحة الانترنت التى تكتب فيها الاكواد

  • @PriyanshuBH-q9w
    @PriyanshuBH-q9w ปีที่แล้ว

    Hi there, I plan on using the EasyOCR Library for some sensitive Documents, Is it safe, like can any data Leaks Occur, also Is there any Documentations of the Library I can refer to ?
    Thanks !!

    • @muskanrath7125
      @muskanrath7125 7 หลายเดือนก่อน

      Same question. Any one please answer?

  • @shooby117
    @shooby117 ปีที่แล้ว

    I run into the following error when I try langchain's UnstructuredImageLoader:
    TypeError: stat: path should be string, bytes, os.PathLike or integer, not JpegImageFile

    • @rashmikasaha2874
      @rashmikasaha2874 7 หลายเดือนก่อน

      did you find a fix to this?

  • @anubhav963
    @anubhav963 ปีที่แล้ว +1

    I am getting list index of range for langchain. Can you suggest sometime there

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      Can you please share the complete error you are getting?

    • @anubhav963
      @anubhav963 ปีที่แล้ว +1

      Sure

    • @anubhav963
      @anubhav963 ปีที่แล้ว

      IndexError : list index out of range .
      For the part where I am trying to get multiple pages converted pdf and and do data[index].page_content
      I am getting an error for this line .

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      @@anubhav963 please let me know if you need further help!

  • @piyushchhawachharia_0068
    @piyushchhawachharia_0068 9 หลายเดือนก่อน

    I used a third party tool API to extract text and tables but the image part is not working for it. It’s not even recognising the images. If I just use the python libraries instead that will recognise the image and I can save them to other folders and later work on it but the extraction of tables won’t work I guess for python libraries.

    • @ajarivas72
      @ajarivas72 9 หลายเดือนก่อน

      I use a neural network to extract data from tables .

  • @Mr_whi
    @Mr_whi 2 หลายเดือนก่อน +1

    Hi sir i am from india and this is very useful i am also doing same kind project…i need your help can you help please?

    • @techwithzoum
      @techwithzoum  2 หลายเดือนก่อน

      I am glad it helps. How may I help?

    • @ChummaSmith
      @ChummaSmith หลายเดือนก่อน

      Hey, I am also from India, What kind of project are you working on?

  • @ibrahimkouma6751
    @ibrahimkouma6751 ปีที่แล้ว +1

    Bonjour êtes vous malien? super tutoriel merci pour le partage

    • @techwithzoum
      @techwithzoum  ปีที่แล้ว

      Je suis Ivoirien avec des originines Maliennes!

    • @ibrahimkouma6751
      @ibrahimkouma6751 ปีที่แล้ว

      @@techwithzoum Je suis un Malien .Y'a t'il un moyen de vous contacter, je dois commencé mon programme de master en data science et machine learning bientôt plus précisément en septembre. j'ai bésion de poser des questions à quelqu'un qui est déjà dans l'industrie.

  • @QorQar
    @QorQar ปีที่แล้ว

    كل الاساليب لتحويل الملفات فشلت ولا اعرف السبب

  • @valmirrastelyjunior9400
    @valmirrastelyjunior9400 10 หลายเดือนก่อน

    Ok

  • @hajarabdullah4648
    @hajarabdullah4648 8 หลายเดือนก่อน

    that was very useful for me, but i faced one struggle in langchain unstructuredimageloader it say its not allowed PNGImage? i don't get it, I try to resolve it but i couldn't , pls if there is any way to contact you, I'll appreciate . thank you