Contact Us

TEL: +86-020-82497630
Mob: +8613694265790
E-mail: ritachen@flythinking.com
Address: 1F,2F Building 8 Yujing Industry Zone, Dalingshan Road, Tianhe District, Guangzhou, Guangdong, China

Home > News > Content
MIT CSAIL Use AI From 2D Video Create A 3D Printing Sports Sculpture
Sep 20, 2018

Understanding of the movement is the basis of all living species, whether it is in the calculation of the pitch Angle, or see the movement of the predator and prey.But simple video actually cannot give us a comprehensive understanding.Because used to study the movement of the traditional video and photo is two-dimensional, so they don't show us people who are interested in or subject the potential of 3D structure.


Now, computer science and artificial intelligence laboratory at the Massachusetts institute of technology, Google research institute and researchers at the university of California, Berkeley, have put forward a kind of artificial intelligence (AI) system to show how the human body movement.

The system is called MoSculp, it USES an algorithm can take 2D video and transform it into 3D printing "sculpture" movement.It can be for professional athletes, dancers, or anyone who want to improve physical skills to provide more detailed movement research.

"Imagine that you have a video about Roger federer in the tennis match pitches, and a video of yourself to learn tennis," doctoral Zhang Xiu confessed."Then you can set up two scenarios of the sculpture to compare them, and more fully study where you need to improve."


Here is its working principle: the video is loaded into the system, MoSculp will detect the key cover on the input pin, and through several randomly selected frame to confirm them.(the built-in calibration tool allows users to adjust when necessary.)After adjustment "temporary inconsistent detection", it will generate movement sculpture and loads it into a custom interface.

This is a multi-step process, all MoSculp demand is video sequence.Here is how it works: the video is loaded into the system, MoSculp will automatically detect the subject first 2D key points on the body, for example, the ballerina hip, knee and ankle, simultaneously complex dance sequences.It will then these points in the best position into 3D "skeleton".