Arcade game highscore recognition

A neural network based program to detect whether or not a highscores has been reached

I will try to be building a python program to detect the current game played on an arcade machine,web scrape highcores from opfinderklubben.dk and check if a highscore has been reached

MAIN FILE: main.py

LOG

Friday November 1st 2019

• Predictions are now more accurate thanks to noise when creating the training data. this reduces the perfectness of the images. making it easier to predict real life images
• New files
  • load_frames_from_video.py
    • Extracts frame from a gameplay video of a given game and apllies noise, rotation and cropping. to make it less perfect
  • create_training_data.py
    • Takes the outputted images from load_frames_from_video and turns them into training sets (pickle files)
  • create_model.py
    • Takes the training data from create_training_data and trains a fully connected neural network
  • test_model_with_webcam.py
    • Opens the webcam and makes prediction based on the frames
• New folders
  • models
  • training_sets
  • files_not_in_use
    • for files not in use

Saturday October 19th 2019

• added model for:
  • Frogger
  • Donkey Kong
  • Amidar
• creted tflite model for rasperry pi

Friday October 18th 2019

• i have now trained a neural network using python and is able to classify frogger and donkey kon with an acurracy of 100% during training