Skip to content
A quick implementation of MLFLIP
C++ Python C CMake
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
dependencies/zlib-1.2.8
resources
scenes
source init Apr 18, 2019
tools
CMakeLists.txt
INSTALLING
LICENSE
README.md init Apr 18, 2019

README.md

MLFLIP

You can find the source codes for the SCA paper: [Um et al., 2018, "Liquid Splash Modeling with Neural Networks"].

Requirements

Activate your tensorflow environment on your machine; e.g.,

source ~/tensorflow/bin/activate

You need to compile the delivered mantaflow sources. Please refer to the general guideline to compile mantaflow: http://mantaflow.com/install.html

MLFLIP requires a special cmake option for numpy support; e.g., in your build directory,

cmake .. -DGUI=ON -DNUMPY=ON

Run training simulations

for i in {00..09}; do
  ./manta ../scenes/tsim_flip.py --nogui --seed=$i -o /tmp/tsim_flip_$i
done

Generate training data

for i in {00..09}; do
  ./manta ../scenes/tdata_gen.py -o /tmp/tdata/tsim_flip_$i /tmp/tsim_flip_$i
done

Train a model

../scenes/tf_train.py --mve -o /tmp/tfmodel/ /tmp/tdata/

Run MLFLIP

./manta ../scenes/mlflip.py --load /tmp/tfmodel/

If you want to use a pre-trained model, please try:

./manta ../scenes/mlflip.py --load ../scenes/tfmodel/
You can’t perform that action at this time.