Mar 30, 2020 21:00:00

Attempt to automatically recognize the mark that Naruto and Kakashi connect by hand with a neural network with `` NARUTO ''

The manga '

Naruto ' series by Masashi Kishimoto on the theme of a ninja is a globally popular work with a cumulative total of more than 250 million copies published worldwide, and a sequel is serialized at the time of article creation. AngryCoder of YouTuber has developed an algorithm that detects and recognizes in real time a mark tied by hand when such a character of 'NARUTO' performs ninjutsu from the camera.

This Machine Learning Algorithm Detects Naruto Hand Signs-VICE
https://www.vice.com/en_us/article/xgq3d3/this-machine-learning-algorithm-detects-naruto-hand-signs

In 'NARUTO', most of the ninjutsu activation conditions are set to 'sign by hand'. There are 12 types of seals derived from the zodiac, and the characters in the play use various techniques depending on the combination.

AngryCoder decided to pursue a project based on his favorite TV anime version of `` NARUTO '' when studying machine learning, and when he signed the signs appearing in the play in front of the camera, We created an algorithm that automatically recognizes the mark and specifies the type of mark. AngryCoder commented, 'I didn't want to do projects related to analysis and statistics. Mathematics is pretty boring.'

You can see how the algorithm developed by AngryCoder actually works from around 1 minute 29 seconds of the following movie.

Used Deep Learning To Detect Naruto (Anime) Hand Signs-YouTube


On the right side of the image are the twelve signs that appear in the TV anime version.

When you put your hands together in front of the camera, Rooster ( Bird ) Was recognized as a sign.

Put your left hand on the right fist Dog ( Dog ) Mark

Align your palms with your fingers Dragon ( Tatsu ) The mark has already been judged before the sign is finished. The response speed and accuracy of the algorithm are considerable, making it a perfect algorithm for NARUTO fans to practice signing.

AngryCoder said he built a convolutional neural network using Keras , an open source neural network library, based on the single-layer neural network, and performed machine learning using the data set of signs that appeared in the play. The input image is converted to a character string of 1 and 0, and through a trained algorithm, it is determined by probabilistically evaluating which mark is connected.

AngryCoder commented, 'Some seals were really hard to tie. I'm not a ninja!'