freshlybuiltimagebol

Image to audio file python libary in development

freshlybuiltimagebol is the library made to allow users to perform various operations with text and images including translation to different languages, generarting audio file from text,extracting text from image and many more.

Installation

To use the library just run the command pip install freshlybuiltimagebol from your terminal. If the library is already installed than try pip install --upgrade freshlybuiltimagebol.

If any occurs persists then reinstalling the library might help, try pip install --force-reinstall freshlybuiltimagebol.

Features

1. Translating and generating audio files from text

a) Translating Text This is used to translate text in different languages. Run freshlybuiltimagebol.ShabdDhwani.shabd_ki_bhasha_badlo(shabd,bhasha). Here 'shabd' is the text you want to translate and bhasha is the language in which you want to translate. Use lowercase in language name.

Example freshlybuiltimagebol.ShabdDhwani.shabd_ki_bhasha_badlo("Welcome to Freshlybuilt","hindi")

b) Pronounciation of the translated text We can also get a glimpse of how to prnounce the translated text. Run freshlybuiltimagebol.ShabdDhwani.bhasa_badlkr_kya_bole(shabd,bhasha).

c) Know the language of text (shabd_ki_bhasha_jaano) Run freshlybuiltimagebol.ShabdDhwani.shabd_ki_bhasha_jaano(shabd).

d)Convert text to sound output(shabd_se_dhwani) Run freshlybuiltimagebol.ShabdDhwani.shabd_se_dhwani(shabd,bhasha,filename).

e)Recognize text from natural scene image(natural_photo_se_text) Run freshlybuiltimagebol.imagebol_model_downloader('F_est') first to download the model, Right now, there is only one available model i.e 'F_est'. Run freshlybuiltimagebol.NaturalPhotoShabd.text_pehchano(image).

freshlybuiltimagebol.NaturalPhotoShabd.text_pehchano(image,min_confidence=0.85,width=320,height=320,padding=0.00)

Parameters:

• image=input image (type=numpy.ndarray) -required
• min_confidence= minimum confidence threshold to detect text from image (default=0.85) -optional
• width=resizing width of image (default=320) -optional
• height=resizing height of image (default=320) -optional
• padding = padding of text box around detected text (default=0.00) -optional

f)Preprocess image before text extraction(OCR_Printed_text.py) Run

`freshlybuiltimagebol.ImageProcess.remove_noise(image)`.  
`freshlybuiltimagebol.ImageProcess.remove_skew(image)`.  
`freshlybuiltimagebol.ImageProcess.sharpness_blur(image)`.  
`freshlybuiltimagebol.ImageProcess.to_text(image)`.  
`freshlybuiltimagebol.ImageProcess.plot_image(image)`.  

Parameter:
image=input image (type=numpy.ndarray) -required

e)Detect Page and crop image(page.py)
Run

`freshlybuiltimagebol.Page.detect(image)`.    

Parameter:
image=input image (type=numpy.ndarray) -required

imagebol_model_downloader(downloader.py) usecase - to download pretrained models to local machine during implementation for the first time.

method - from freshlybuiltimagebol import imagebol_model_downloader(model_name)

parameters:

• model_name = model name defined in the dictionary structure of available model inside imagebol_model_downloader class

updates in downloader needed to be done while adding new models dictionary structure-

available_models={ "F_est":["frozen_east_text_detection","94.4MB","8a9b7f2ebd9bcf8212bfa856b065e6f0"] }

• here key "F_est" is the model_name used as the parameter
• "frozen_east_text_detection - actual file name (extension should be .pb)
• "94.4MB" - actual file size
• "8a9b7f2ebd9bcf8212bfa856b065e6f0" hash key generated using md5 encryption technique

important while updating model file , hash key must be updated.

Contributors

Vishal Sharma: G.B. Pant Govt. Engineering College, New Delhi

Kapil Bansal: G.B. Pant Govt. Engineering College, New Delhi

Ravi Pawar: G.B. Pant Govt. Engineering College, New Delhi

Sushmita: G.B. Pant Govt. Engineering College, New Delhi

Simran: Maharaja Agrasen Institute of Technology, New Delhi