Skip to content

Latest commit

 

History

History
48 lines (47 loc) · 1.2 KB

Training.md

File metadata and controls

48 lines (47 loc) · 1.2 KB

Install Dependencies

  • Install python dependencies
sudo apt-get -y install python2.7-dev
sudo apt-get install libhdf5-dev
cd experiments
pip install --user -r requirements.txt
  • Install the deepmind/torch-hdf5 which gives HDF5 bindings for Torch:
luarocks install https://raw.githubusercontent.com/deepmind/torch-hdf5/master/hdf5-0-0.rockspec

Download and Prepare the Data

  • Download the dataset
wget http://msvocds.blob.core.windows.net/coco2014/train2014.zip
wget http://msvocds.blob.core.windows.net/coco2014/val2014.zip
unzip train2014.zip
unzip val2014.zip
  • Create HDF5 file
python make_style_dataset.py \
  --train_dir train2014 \
  --val_dir val2014 \
  --output_file data.h5
  • Download the VGG-16 Torch model
bash models/download_models.sh

Train the Model

  • Training a 16 style model. For customized styles, set style_image_folder pointing at the folders containing the style images.
th main.lua \
-h5_file data.h5 \
-style_image_folder images/9styles \
-style_image_size 512 \
-checkpoint_name 9styles \
-gpu 0
  • Test the model
th test.lua \
-premodel 9styles.t7 \
-input_image images/content/venice-boat.jpg