Although I really enjoy learning, certain topic could be a long and arduous journey which often land me scrolling through the internet looking at dogs and cats or random youtube music videos. Thus, instead of researching or learning a computational topic in depth today, the designer in me decided to write some cheap quick and easy code but able to get something appear on my screen just for the cheap thrill of feeling like a “CoMpUtAtIoNaL DeSiGnEr”.
Beside being a computational designer, I am also a recent practitioner of yoga. While trying out new and more challenging poses, I often find the process of balancing as something iterative. Once finding the balance, the right spot, the pose become much easier and require much less strength. Inspired by the practice, I been trying to bring images of yoga poses photo to a 3D vector skeleton in Rhino.
I run the yoga pose image through two algorithm: poseNet and denseDepth. poseNet returns the human skeleton, with the “coordinates” of the joints. These coordinates are then mapped onto denseDepth image, return us the “brightness” of that pixels. All these numbers are then brought into Rhino and make up a 3D human skeleton.
Original Image
PoseNet
Dense Depth
Dense Depth with mapped joints (zoom in closely)
3D result…sorry for the exploded head hehe.
I understand that there are much better algorithm to do this such as densepose and the result looks very different from the image, but hey this is just an attempt for me to have some fun!