Handwriting to Text Conversion using Deep Learning
ฝัง
- เผยแพร่เมื่อ 19 ก.พ. 2021
- #datascience #deeplearning #machinelearning
In this webinar we will look at how we can extract content out of handwritten documents into a readable text format. This demo will be using Time Distributed CNN along with LSTM and CTC Loss Function
Below is topic that will be covered in the session
Project and Business background
Annotating the Images
Model Architecture and Training
Challenges Faced
Demo
Q&A - บันเทิง
I am not exaggerating.. You are a pride of our India, on your own terms..Great such high-quality content in a noisy and shallow internet world .. Wonderful job
Much awaited session ! ! ! Finally you delievered it Thank you Sir and team
Very interesting, providing complete timeline and full reference Thankyou Sir.
Very interesting project.
Excellent & knowledgeable session.
Hi srivatsan. Great session by girish and balaji. @srivatsan Can you please share the keras ocr finetuning notebook or any other reference for same.. thanks a lot.
Here is the link to git repo for code - github.com/GireeshS22/TimeDistributed-CRNN
Could you please demonstrate the running of this code
Hi Giresh can you please suggest hiw can we remove pencil scribbles from na scanned page?
How to get image -word data from the vgg software. For training the model we required image and corresponding word. You annotated in vgg but wheather we get this pair as output like IAM dataset.
nice idea...can I get training data, so I can experiment further.
thanku sir
Can you please provide annotated data for handwritten text recognition of doctors prescription??? It will be a great help.....Please do reply
When i run the Model,py you sent it gives an error . Cannot convert a symbolic Tensor (bidirectional_2/forward_lstm_2/strided_slice:0) to a numpy array. This error may indicate that you're trying to pass a Tensor to a NumPy call, which is not supported . Can you guide me how to solve this?
Thank you for presenting this project, very interesting. I am trying to build an OCR model for Gujarati language. I've gathered enough data(structured data). The existing datasets and models for Gujarati do not give accurate results on old books, so I am trying to create a dataset for them. I need some guidance on creating annotated dataset. Do you have an email ID, so I can reach out to you.
Hello, this is really good content. Do you have any content on using computer vision to identify manipulated invoices?
Neil..By manipulated you mean identifying fraudulent invoices right. We do not have content but I have done some similar work using autoencoders combined with features extracted from invoices
@@AIEngineeringLife hello Srivatsan, yes by manipulation I mean people using photo editing tools or adobe acrobat to change text within invoices. I’ll read up the topics you mentioned and it would be great if you could do a tutorial on this topic. Thank you
It was mentioned during the session that links to certain research articles will be provided. Could you please advise where I can find them?
This is the one Naveen - arxiv.org/abs/1811.07768
If you are looking for video specific to OCR you can check this playlist videos - th-cam.com/play/PL3N9eeOlCrP4uLCtas5vxq09sWz6jJXrw.html
Do you have signature detection material?
can you please provide a annotation process for text recognization?
Can i get soure code?
Sir can you please share the code?
bro How to collect dataset for this project
Can you share the code please!
Hi I have a project can I ask you? Do you have an email ID? Also, I tried to run your code on my handwritten image but it didn't work...Thanks
hey bro can we connect over email i need some of ur help... regarding this kinda project
please python code