tmurph / datasci-smash

A toy demo for some techniques in machine vision and machine learning

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Author:Trevor Murphy

This repository contains my notes and work applying neural networks to footage from the Super Smash Bros. Melee video game.

So far, I have used transfer learning techniques to train a VGG16-based neural network to classify images of single characters. See the classification notebook for details.

Next, I will train a network to locate a character within an image. Then locate and classify two characters in the same image.

After that, I will train a network to recognize the location and combat move of the characters within the image.

Finally, I hope to make these networks fast enough to run concurrently with streaming video sessions, to enable real time analysis of matches in progress.

About

A toy demo for some techniques in machine vision and machine learning

License:GNU General Public License v3.0


Languages

Language:Jupyter Notebook 97.9%Language:Python 1.0%Language:Makefile 0.9%Language:Shell 0.2%