synthetic_data_readme.md

Character Image Generation

This repository contains a script to generate synthetic images of characters using various fonts. The images are augmented with random transformations such as rotation, shear, and zoom.

Requirements

Ensure you have the following packages installed:

• Pillow
• concurrent.futures
• multiprocessing

You can install the required packages using pip:

pip install Pillow

Directory Structure

├── fonts/           # Directory containing .ttf font files
├── images/
│   └── dataset/     # Directory where generated images will be saved
└── generate_images.py  # This script

One important thing to note is that this script stores the images in it inside a folder with it's ascii number.

For example

[A] => 65/
[B] => 66/
[C] => 67/
[D] => 68/
...
..
.

and so on.

Character Range

This is the default character range.

! ,  # ,  $ ,  % ,  & ,  ( ,  ) ,  * ,  + ,  , ,  - ,  . ,  / ,  0 ,  1 ,  2 ,  3 ,  4 ,  5 ,  6 ,  7 ,  8 ,  9 ,  : ,  ; ,  < ,  = ,  > ,  ? ,  @ ,  A ,  B ,  C ,  D ,  E ,  F ,  G ,  H ,  I ,  J ,  K ,  L ,  M ,  N ,  O ,  P ,  Q ,  R ,  S ,  T ,  U ,  V ,  W ,  X ,  Y ,  Z ,  [ ,  ] ,  ^ ,  _ ,  a ,  b ,  c ,  d ,  e ,  f ,  g ,  h ,  i ,  j ,  k ,  l ,  m ,  n ,  o ,  p ,  q ,  r ,  s ,  t ,  u ,  v ,  w ,  x ,  y ,  z ,  { ,  | ,  }

by default the size of generation is set to 2400 images per class.