spawnwvinull/weseeyou repository overview

⁠The WeSeeYou machine learning art project

This a dockerized machine learning art project based on the neural style transfer principles described by Leon A. Gatys, Alexander S. Ecker, and Matthias Bethge in their paper 'A Neural Algorithm of Artistic style'⁠ first published in 2015.

See the official project documentation here⁠.

⁠Base Image Tags

-The images are based on the tensorflow images⁠ on this docker hub. Currently we are using tag 2.2.1-gpu⁠

• aimachine-* is the image for image style transfer. It has both the production as the research components on board.
• ui-* is the backend management interface. Via this ui an overview of the available styles, created images and connected drones is available.
• drone-* is the image for hooking up simple display enabled devices for displaying the output of the aimachine.
• Use latest for production purposes and latest-dev for development purposes.

⁠Run containers

To run a container you could use the following:

$ docker run -it --rm spawnwvinull/weseeyou:aimachine-latest /bin/bash

The images are based on the tensorflow-gpu enabled version which means the the --gpus all flag is available.

$ docker run -it --rm --gpus all spawnwvinull/weseeyou:aimachine-latest /bin/bash

⁠AIMachine volumes

Inside the image there is an input and output volume available that can be used externally with a mounted volume.

-v 'your_path'/input:/app/input
-v 'your_path'/output:/app/output

And optionally you can provide an external log path for the log files.

-v 'your_path'/logs:/app/logs

⁠AIMachine research

If you want to use the application in research mode to analyse style impact you can specify an alternative start command

cmd ./startResearch.sh

You can also specify volumes to get direct access to the research results.

-v 'your_path'/output:/app/Research/output
-v 'your_path'/output:/app/output

And optionally you can provide an external log path for the log files.

-v 'your_path'/logs:/app/logs

⁠Docker compose

You can use docker-compose to orchstrate the applications. A typical setup could start an aimachine and an aimachine in research mode together with a ui component. Keep in mind that providing a GPU to a container in a docker-compose file is done via the environment variable syntax.

environment:
      - NVIDIA_VISIBLE_DEVICES=1

Author: Jeroen Vesseur

Website: http://www.wvinull.com/⁠

License: GNU General Public License⁠