soapisnotfat/experience-extraction-from-gameplay-video

Player Experience Extraction from Gameplay Video

Official code release for AIIDE 2018 oral presentation paper Player Experience Extraction from Gameplay Video, authored by Zijin Luo, Matthew Guzdial, Nicholas Liao and Mark Riedl.

Table of Contents

• Getting Started
• License
• Contributing

Getting Started

This is the repo for the skyrim dataset experiment.

For the Gwario experiment, please checkout the gwario branch

Prerequisites

• PyTorch ~0.4.0
• torchvision ~0.3.0
• numpy
• PIL
• tqdm

Installation

git clone https://github.com/IvoryCandy/experience-extraction-from-gameplay-video.git

Structure

- data
  - dataloader.py
  - download.sh
- model
  - model.py
  - download.sh
- main.py
- misc.py
- solver.py

main.py has an argument setting, you can run set your custom hyper-parameters here

solver.py has three solvers -- Backprop, ImageNet, and TeacherStudent. Each solver contains all reletive functions for loading data, training model, and tesing performance. Use them based on your purposes.

Usage

to download data

cd data
bash download.sh

to download pretrained model

cd models
bash download.sh

run under default config

python3 main.py

For customized experiments, select different solvers from solver.py and put it in main.py, and set your own hyper-parameters in arguments.

License

Distributed under the Apache2 license. See LICENSE for more information.

Contributing

• Fork it