In our paper we describe a faster way to generate textures and stylize images. It requires learning a feedforward generator with a loss function proposed by Gatys et. al.. When the model is trained, a texture sample or stylized image any size can be generated instantly.
Instance Normalization: The Missing Ingredient for Fast Stylization presents a better architectural design for the generator network. By switching batch_norm to instance norm we facilitate the learning process resulting in much better quality.
Download VGG-19.
cd data/pretrained && bash download_models.sh && cd ../..
Basic example:
th train.lua -data <path to any image dataset> -style_image path/to/img.jpg
The image dataset should be structured as in fb.resnet.torch having train and val folders and some folders corresponding to classes as you were doing classification. You can create a dummy folder train/dymmy/ and val/dummy/ and store all the images in them. Only images from train forlder will be used. Change the code or rename folders to use val folder. You can use any dataset for example mscoco or imagenet. Use validation part if using imagenet.
To achieve the results from the paper you need to play with -image_size, -style_size, -style_layers, -content_layers, -style_weight.
Do not hesitate to set batch_size to one, but remember the larger batch_size the larger learning_rate you can use.
th test.lua -input_image path/to/image.jpg -model data/checkpoints/model.t7
Play with -image_size here.
You can find a pretrained model here. It is not the model from the paper.
soon
- The code was tested with 12GB NVIDIA Titan X GPU and Ubuntu 14.04.
- You may decrease
batch_size,image_sizeif the model do not fit your GPU memory. - The pretrained models do not need much memory to sample.
The code is based on Justin Johnson's great code for artistic style.
The work was supported by Yandex and Skoltech.