Center for Technology and Art, National Tsing Hua University - NTHU, (TW); Department of New Media Art, Taipei National University of the Arts - TNUA, (TW)
Impressionism expressed the scientific meaning of "light" in artistic creation. The light color reflected by the luminous flux through the medium in the air became the concept of impressionists’ creation.
After collecting the web imagination of global citizens about COVID_19, we trained the generative model and let the computer randomly synthesize it. We adjusted various parameters during the training process making the results such as Dou Jia, Monet, Renoir , their style paintings.
Through this work the diffusion path of COVID_19 spreading the molecular of the virus through the air, just like the scattering phenomenon formed by light through the particles in the air. Color is exclusive to the humans’ world, and COVID_19 is also born with humans.
The colorful world we see in this work represents the equality of races on the earth. The array of 196 pictures represents 196 countries were invaded by the COVID_19. As human beings, none of us are spared from this disaster.
https://techart-ars.tw/2021/?page_id=293
https://www.facebook.com/permalink.php?story_fbid=2944347522470150&id=1944000092504903
python -m pip install -r requirements.txt
This code was tested with python 3.7
This script is run at DGX-1 with 8 NVIDIA® Tesla® V100. We will use in the training tf.distribute.MirroredStrategy, that supports synchronous distributed training on multiple GPUs on one server. It creates one replica per GPU device. Each variable in the model is mirrored across all the replicas. These variables are kept in sync with each other by applying identical updates. Here is the simplest way of creating MirroredStrategy:
mirrored_strategy = tf.distribute.MirroredStrategy()
We used data parallelism to split the training across multiple GPUs to reduce model training time. Each GPU has a full replica of the neural network model, and the variables are updated synchronously by waiting that each GPU process its batch of data. Please refer to the following:
python main.py
main.py is used to train our model at DGX-1.
GAN.py is our model to learn the style and texture of covid-19 pictures in various countries.
Linz2021.ipynb is in the form of a Jupyter Notebook as a simple display of model training and a Covid-19 image generator.