Anime or Cartoon? — AI-powered style definition now on mobile

Dávid Komorowicz
Jan 12, 2018 · 3 min read
Image for post
Image for post
Photo by Barnabás Neogrády-Kiss

Since last summer I have been working on establishing a new definition for Anime and Cartoon using Artificial Intelligence.

During this time I took part in the KiBu Lift Program and the Hungarian National Talent Program.

Image for post
Image for post
Photo by Barnabás Neogrády-Kiss

At KiBu, Demo Day is a biannual event showcasing the output of the KiBu Talent & Lift programs.

Along with a quick talk about the project I released the first public beta of the app which you can try out here: dawars.me/anime/

The UI

Photo by Barnabás Neogrády-Kiss

The app works much like a traditional camera application. One has to point the mobile at the analysed image and at the top of the screen the app displays the anime or cartoon likeness in percents.

How does it work

The app takes advantage of the superhuman capabilities of neural networks in computer vision, especially CNNs (convolutional neural network).

Image for post
Image for post

The network is trained on hundreds of thousands of images from both anime and cartoon images. Here I would like to stress that only universally accepted animations go into either categories, e.g. Avatar: The Last Airbender is excluded.

During this phase the networks creates an internal representation of “anime likeness” and “cartoon likeness”.

We don’t really know what these are but the network distills these from the shown images. Usually these features can be related to outlines, color usage or bigger shapes like eyes, hair.

The current beta utilizes the Mobilenet architecture, which offers great tradeoff between speed and accuracy. Ideal for real time image classification tasks, just like this project.

Future work

Currently the app does a good job at recognising and distinguishing between anime and cartoon images. A current limitation is that it can only say something is either anime or cartoon but not both (a mix of them) or neither.

I’m planning to train an even bigger Neural Network (VGG-16) on this dataset. VGG-16 is well known for its great performance in Style Transfer.

Using a slightly modified version of Neural Style Transfer we could get a glimpse of the inner workings of the network.

My hypothesis is that if we tell the network to make a cartoon image more anime looking, certain features will appear that the network associates with anime and vice versa.

Image for post
Image for post
Image for post
Image for post

Originally published at dawars.me.

Kitchen Budapest Stories

Collection of articles written by members and peers of…

Dávid Komorowicz

Written by

Machine Learning Engineer interested in Deep Learning and Virtual Reality. dawars.me

Kitchen Budapest Stories

Collection of articles written by members and peers of Kitchen Budapest, the Open Innovation Lab of T-Systems Hungary. Articles about technology, innovation, digital literacy, DIY, research and education.

Dávid Komorowicz

Written by

Machine Learning Engineer interested in Deep Learning and Virtual Reality. dawars.me

Kitchen Budapest Stories

Collection of articles written by members and peers of Kitchen Budapest, the Open Innovation Lab of T-Systems Hungary. Articles about technology, innovation, digital literacy, DIY, research and education.

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store