This was a fun little weekend project I made by smooshing together a few libraries. I was surprised by how fast and accurate the pose estimation library [0] was, especially running real-time in the browser!
This project requires some extensive setup and special hardware like a projector, so I expect most people won't be able to try it on their own -- which is a bummer, because it's really fun to play! But there are videos linked in the readme, and I'll link them again here [1][2]
Great project. Most games are too sedentary and making them physical has the potential to make a healthier genre of games. I’ve had some similar ideas that I never put to practice but the main idea was augment sports games with technology and bring them to the 21st century. Think about a basketball referee that keeps track of rules, bounds, etc or a mixture of augmented ball played with a phisical ball but which could interact with the background: eg projected shapes, sounds; etc
This was fun to play. I set it up today and we all enjoyed it. We noticed that it helps to jump forwards in the beginning until the cacti speed up, otherwise you'll usually land on the cactus. It's much harder to achieve a comparable score than on chrome://dino though. The only obstacles are cacti in this version, but I think ducking under birds would be far too hard.
4 comments
[ 2.9 ms ] story [ 15.0 ms ] threadThis project requires some extensive setup and special hardware like a projector, so I expect most people won't be able to try it on their own -- which is a bummer, because it's really fun to play! But there are videos linked in the readme, and I'll link them again here [1][2]
[0] https://github.com/tensorflow/tfjs-models/tree/master/posene...
[1] https://www.youtube.com/watch?v=T-GrpBywY18 demo
[2] https://www.youtube.com/watch?v=AVGCd5r4ihw calibration sequence