README.md

Launcher for Game

Purpose:

My childhood game requires multiple steps to login and load the game. Moreover, their website is very slow and would take ages to load.

I created this application to login and launch the game easily, as well as to translate anything in the game with a click of a button instead of using Google Translate Image on my phone.

Features

Login Window

• Load login credentials from .env file (will change in the future)
• Solve Captcha Automatically (86% correct)
• Check for the newest server number to validate user input

Game Window

• Custom TitleBar
• Mute/unmute application
• Check for the newest server number to validate user input
• Snipping Tool to capture an area on screen and ultilizes the Image Translator to translate it

Images

• Gif of the current application:
  
  

• Gif of the image translator:
  
  

• Gif of the zoom in/out feature (CTRL + scroll => map's size; CTRL + Shift + scroll => menu size):
  
  

• Some sample images:

  • Loaded login screen:
    
    

  • Loading game screen:
    
    

  • Result of the image translation:
    
    

Stats

Image Translator

• ~0.9 seconds to recognize and extract the text from an image (depending on the size).
• ~0.2 seconds to bulk translate all of the text
• 200MB-4GB RAM usage depending on the image size and the number of text on image.

Captcha

• ~0.3 seconds to solve each captcha.
• 86% accuracy from 100 randomly chosen images

Image Translator

For more information, check out https://github.com/pat266/Image-Translator. This repo shows different processes which showcase their results (accuracy), computation time, and memory usage

Current process:

• EmguCV + pre-trained Onnx model
• Google Translate API (free) to translate it to English

Some other failed text extraction processes

• Capture2Text (extracting text)
  • Inspired by https://github.com/phatjkk/DragonTranslator
  • http://capture2text.sourceforge.net/#text_line_capture
  • ./Capture2Text_CLI.exe -l "Chinese - Simplified" --trim-capture --line-breaks -i "testImg/sample.PNG" -o capture2text_result.txt
  • The result is not very good. Many Chinese characters were not recognized.
• IronOCR-Chinese to detect the Chinese characters
  • Even though it is much better than Tesseract, it was still not able to detect the text on screen at a satisfactory level
  • It is also slow
• EmguCV alone (https://www.youtube.com/watch?v=KHes5M7zpGg)
  • Is worse than IronOCR
  • Does not work well to detect text in game. Works best when the background is clear.