trainsn / insitu_net

InSituNet: Deep Image Synthesis for Parameter Space Exploration of Ensemble Simulations (SciVis 2019)

InSituNet

PyTorch implementation of the deep learning model introduced in our SciVis 2019 paper "InSituNet: Deep Image Synthesis for Parameter Space Exploration of Ensemble Simulations".

Comparison of InSituNet trained with different loss functions

Example visualization images generated by InSituNet

The way of training

unzip mpas_sub.zip 
mv mpas_sub/train/params_sub.npy mpas_sub/train/params.npy
mv mpas_sub/test/params_sub.npy mpas_sub/test/params.npy
mv mpas_sub datasets
cd model 
python main.py --root ../datasets --dsp 1 --gan-loss vanilla --gan-loss-weight 1e-2

More MPAS-Ocean Images

You may find more MPAS-Ocean Images here.

The way of evaluation

python eval.py --root ../datasets \
               --dsp 1 \
               --resume {the path to the saved tar model} \
               --id {image id in the testing dataset}

Then the ground truth, predicted, and difference images are saved.