UPDATED_MASK_RCN_INSTRUCTIONS.txt

conda create --name MASKRCNN python=3
conda activate MASKRCNN
conda install jupyter
conda install matplotlib
conda install tensorflow-gpu
conda install scikit-image
conda install scikit-learn
conda install keras

git clone git@github.com:VCG/perception.git ### replace with your fork
cd perception
git submodule init
git submodule update

jupyter notebook
# have jupyter notebook configured according to https://cs410.net/notes/01/

# now on your local machine
# create tunnel (instead of monster use jesse's machine and instead of 44321 use your own port)
ssh haehn@monster.verymad.net -L 44321:localhost:44321 
ssh viper.verymad.net -L 44321:localhost:44321

# use browser to access https://localhost:44321
# and open ipy/ImageTest.ipynb

####
# now for the maskrcnn pipeline
#

# we need the VGG19 from the Haehn 2018 paper:
#
mkdir -p tensorflow.js/models/VGG19_angle_full_variation/
wget -O tensorflow.js/models/VGG19_angle_full_variation/VGG19.zip https://www.dropbox.com/sh/cavjnb91vbwh2n9/AAA6jmH5uA1rP9D9OKRr8QJ4a/RESULTS_FROM_SCRATCH/C.Figure1.angle/2/VGG19?dl=1 
cd tensorflow.js/models/VGG19_angle_full_variation/
unzip VGG19.zip

# open ipy/MASKRCNN.ipynb and it all works :)