DLFR

This repository is for a different languages font recognition via Raspberry Pi project.

Requirement

google-images-download >= 2.4.2
Keras >= 2.2.4
numpy >= 1.15.2
Pillow >= 5.3.0
pyparsing >= 2.2.2
scikit-learn >= 0.20.0
tensorflow >= 1.10.1
imutils == 0.5.1
opencv >= 3.4.4

How to navigate this repository?

You have the three main scripts to run the project. The predict.py used to predict language from the given image, and both the other file use live-stream video feeds to make a prediction (predictPC for laptops and predictRPi for Raspberry Pi.) The model folder has all the results of the trained models on the datasets. Also, the visualization and image pre-processing folder shows how to visualize the model and how to manipulate the images. Note that I did not include the datasets but you can make your own as suggested below.

There are three different version of datasets used in this project:

Version 1: random images that have a specified text init with a different environment. Used open library by hardikvasa called google-images-download. It's a python script that allows you to search and download hundreds of images (https://github.com/hardikvasa/google-images-download) You can use different keywords to specify the size, the type, and the images format.

      $ for example : $googleimagesdownload --keywords "playground" --limit 20 --color red

Version 2: random images of the specified language alone with white background. Created the dataset using Word document of random words taken from different articles and news. Then use an online converter to convert the PDF to separated images.

Version 3: a modified version of version 2 using Pillow python library. Function 1 to do several transformations on the images, e.g., rotation, flipping position and color changing

How to run the code?

Almost every python script in this project has parse arguments you can check before running the code. For example to run the project on Raspberry Pi using SNN model (also noting that for SNN use -w 32 -e32 -f 1 and for VGG -w 64 -e 64 -f -1)

$ python predict.py --model location/simple_nn1.model --label-bin location/simple_nn_lb1.pickle -w 32 -e 32 -f 1

Website

Please check out the project website,https://sites.google.com/stevens.edu/font-recognition/. Also, please check out the Youtube video about the project, https://www.youtube.com/watch?feature=youtu.be&v=iBfhAOs8fxo.

Contribute

Feel free to fork this repository and make your changes.