judyYan / lunania-ai

our first ai.

lunania-ai

our first ai.

install brew

/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"

install git

yum install git

pyenv setup

git clone https://github.com/yyuu/pyenv.git ~/.pyenv
echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bash_profile
echo 'export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bash_profile
echo 'eval "$(pyenv init -)"' >> ~/.bash_profile

install python 3.5.2

pyenv install 3.5.2
pyenv global 3.5.2

install keras and tensorflow

pip install keras==1.2.2
pip install tensorflow
pip install pillow
pip install h5py
brew install graphviz
pip install pydot
pip install pydot-ng
pip install pyparsing
pip install matplotlib

excute

at first you can train with 3 method

1. completely train by yourself

cd catvsdog/01_job
python scratch.py

2. use vgg16 base, train top fc only.

cd catvsdog/01_job
python bottleneck.py

3. use vgg16 base, train top fc and the 16th convenience2D.

cd catvsdog/01_job
python finetuning.py

then use weights to predict cat or dog

you can set mode from 1 to 3(1: scratch, 2: bottleneck, 3: finetuning)

and use image param to specify an image.

for example:

cd catvsdog/01_job
python predict.py --mode 3 --image ../99_data/input/1.png