Simple Python Project - OCR + Language Translation + Text to Speech

#OCR #machinelearning #speech

In this video I demo a simple fun python project that can take pictures of any language, convert to English and then play audio of the text. Not only that one can play the audio in any Accent (Indian, Chinese, Russian, Japanese, American etc). In case if someone is not able to understand English, no worries it can translate and play audio in target language be it French or German

This project can be extended into use cases for social good like

Tourist going to an country and not understanding sign board in a different language or aged people with less visibility who are not able to read some words clearly

[stylohunk]

Make a video on how to convert it into a deployed application where we can take image from camera and then it converts that to speech, so that it will be a nice project for resume.

[subramani]

Hi Srinivasan, very detailed and clear explanation.

[evangelostzimopoulos]

Great project Srivatsan. Quick question - is there a github repository for your projects, do you tend to share these?

[Behindthewheelwithsyed]

Thank you for the video. Around 6:06 you are removing bounding boxes and printing text. The code to do this looks similar to adding bounding boxes. Kindly check if this is the complete code
text_list = reader.readtext('mk1jx.jpg', add_margin=0.55, width_ths=0.7, link_threshold=0.8, decoder='beamsearch', blocklist='=.')
text_list

[emailvarun]

Hi Sir...thank you for this amazing explanation again...
Just wanted to know what if we want to initialise the ocr Reader in such a way that it takes any language instead of a specific language (like tamil 'ta' you used in this video)

[subramani]

For some reason easy Ocr is giving wrong results when I try. It not capturing Tamil. Capturing some jumbled English letters and language is "Hr", when try to detect. Any settings to be done or configurations to be modified?

[dhanshriborse]

Hello sir, Without training any dataset how can you convert that ? plesae explain.

[jagadeeshnaidu]

Hii sir, plz check the code that you uploaded it showing error i took this as my core project but code is not working i think due version difference could you plz verify and reupload plz

[ANUJITHMV-sdfv]

There is a runtime error in ' reader= easyocr.reader(['ta']) '..
How to solve it. ?

[hemanthyelahanka]

Great work and helpful video, ....I am the person checking for the same and helped me a lot....Let me ask you a question do we have option to convert the subtitles of any video to suitable languages?

Can u help me out in this?

[vailantinafernandes]

Sir actually I was getting error using googletrans so I used google_trans_new so that error got solved but I am getting wrong output "India's first sentimental preparation created without chemical combinations of chemical combinations", Sir could you please help me with this.

[saumyautb]

Thanks for this video, with nice and simple explanation of each step. I am a new-entrant in data science field just started learning the concepts.. this video shows: image-ocr-translate- text to speech; how can we go for speech to text ? I have explored a bit of Watson Studio speech to text instance! Unable to figure out the process!

[elangoshunmugaraj]

Thank you, As mention here, Please I will use your project and video for my learning and documentation.

[shivampurbia]

You didn't accept my invite on LinkedIn so texting here, I am planning to modify this and deploy on gcp, should I use streamlit or react for the web part. ?

[mystery_official]

bounds = reader.readtext('mk1jX (2).jpg')
bounds
above lines showing this error pls help
error: Unknown C++ exception from OpenCV code

[sriharisht]

code = ', '')
AttributeError: 'NoneType' object has no attribute 'group'

i got this error what to do sir

[youcefahmanache]

Hi, is returning a 'NoneType' object has no attribute 'group' what do you think is the reason ?

[CodeVibes]

Sir can you make a video on SPEECH TO TEXT as it's there for my final year project.. Please sir

[aishwaryabaskaran]

hi sir how do u take a live capture for the image and translate instead of a saved image
?

[joalar]

Amazing work, just that i need, but i have a questions and a little fixed to work in google colab.

The first question is that i have problems to translate if the PDF have 2 text columns, cause read boot colums in the same row. Any idea to fixed this??

And the little fixed it's:

Original----text_en= translator.translate(text_comb, src='ta')
Fixed ---- text_en = translator.translate(text_comb, lang_src=ta, lang_tgt=en)

Thanks for a great work, i will try to find this many weeks ago.