Tweet
Basically: automatic armature generation from 2D data
Paper.....
New Scientist....(video on page)
Paper.....
New Scientist....(video on page)
Animated 3D models extracted from single-camera video
LOOK away Pixar. There's no need for expensive animation systems to bring characters to life in 3D. You can do it with your own video footage and some software.
Developed by Xiaolin Wei and Jinxiang Chai of Texas A&M University in College Station, the software tracks humans appearing in video footage, and creates 3D models of them.
To achieve this for, say, a video sequence of a weightlifter, the animator marks the position of the person's joints and the directions in which their limbs are pointing, along with points of contact with the weights bar, in several key frames of the sequence. Using this information, an algorithm estimates the person's size and creates a 3D reconstruction for each pose that was annotated.
To make the model move smoothly and naturally, other algorithms track the movements of joints from pose to pose calculating how each one must have moved between those frames. This is used to generate an animated 3D figure which can be viewed from multiple angles. The software can animate a variety of actions, such as walking, jumping, and even complicated acrobatics.
The system could be incorporated into animation software packages and would be useful for amateurs and professionals alike, says Wei. The researchers will present their work at the computer graphics conference Siggraph in Los Angeles later this month.
LOOK away Pixar. There's no need for expensive animation systems to bring characters to life in 3D. You can do it with your own video footage and some software.
Developed by Xiaolin Wei and Jinxiang Chai of Texas A&M University in College Station, the software tracks humans appearing in video footage, and creates 3D models of them.
To achieve this for, say, a video sequence of a weightlifter, the animator marks the position of the person's joints and the directions in which their limbs are pointing, along with points of contact with the weights bar, in several key frames of the sequence. Using this information, an algorithm estimates the person's size and creates a 3D reconstruction for each pose that was annotated.
To make the model move smoothly and naturally, other algorithms track the movements of joints from pose to pose calculating how each one must have moved between those frames. This is used to generate an animated 3D figure which can be viewed from multiple angles. The software can animate a variety of actions, such as walking, jumping, and even complicated acrobatics.
The system could be incorporated into animation software packages and would be useful for amateurs and professionals alike, says Wei. The researchers will present their work at the computer graphics conference Siggraph in Los Angeles later this month.
Comment