link:https://editor.p5js.org/yz10444/sketches/6S51qK7gX

The interaction of slicing fruits was inspired by Fruit Ninja.However, I want to interact in the air without touching the screen.

HANDPOSE RECOGNITION

https://youtu.be/3ZrJrZjkMZY

HARVEST

WATERDROP

SEED

$http://milk.co/treachery#/id/i4207763

What type of machine learning models did the creator use?

Skeleton tracking based on the Kinect sensor. It captures the body outline of the audience through infrared sensors and uses a machine learning model to map the human feature points (head, shoulders, elbows, hands, etc.) to the digital space, and transforms them into huge "shadow birds" on the screen. When the audience waves their arms, the shadows will grow huge feathers and fly.

What data might have been used to train the machine learning model?

Thousands of depth sensor photos of people of different heights, weights and clothing at various angles. Manually labeled joint positions. To make the movements look natural, the model may also have referred to a large amount of motion capture data to accurately predict the next position of the body when the viewer moves.

Why did the creator of the project choose to use this machine learning model?

The author hopes that the audience can directly resonate with the artworks without wearing any controllers or headsets, simply by standing in front of the screen. Machine learning enables the computer to "understand" human body language. This installation explores the birth, death and rebirth of human beings. By transforming human figures into birds through machine learning, it creates a surreal and almost sacred "metamorphosis" experience, which would be impossible without high-precision real-time tracking.