kjshim / autofpop

Auto solver for friends pop

Prerequisites

• adb
• python 2.7 and related libraries (See requirements.txt)
• keras (edge version for this commit)
• tensorflow

Setup (on Mac)

Install adb

$ brew install android-platform-tools

Install conda with python 2.7

$ wget https://repo.continuum.io/miniconda/Miniconda-latest-MacOSX-x86_64.sh
$ bash Miniconda-latest-MacOSX-x86_64.sh
$ rm Miniconda-latest-MacOSX-x86_64.sh

Install requirements with conda

$ conda install -y --file requirements.txt

Install TensorFlow

TODO: description

Install Keras

edge version for this commit

TODO: description

TODO: description for setting ~/.keras/keras.json

Usage

• Make model

$ python -m autofpop.make_model # Models are stored at model/

• Run Friends pop
• Start the target stage
• Execute the program like blow

$ python test_friendspop.py

• Repeat
  • Wait for the result
  • Do just like the result
  • Close the result window

Test

$ nosetests

Files to check

• autofpop/andlib.py
• autofpop/ScreenReader.py
• autofpop/recognition.py
• autofpop/friendspop.py
• tests/test_friendspop.py

autofpop's Issues

Improve recognition techniques

As is: Recognize

• Black & normal
• Black & stripe1
• Black & jail
• Blue & normal
• Blue & stripe1
• ...

To be: Recognize

• Color
  • Black
  • Blue
  • ...
• Type
  • Normal
  • Stripe 1
  • Jail
  • ...
• ...
  • ...

Add the travis badge on README.md

The badges are cool.

Check out this as an example: https://github.com/ermaker/honeypot

Add basic test cases

It is hard to modify current code without test cases.