In week 4, we continue to explore on the platform. Here are some new insights we got on OpenPose. The good news is OpenPose is in the perfect range of computational load to demonstrate edge computing. Still, there’s a limitation for that: becauses it’s a stream of images, with xy coordinates of 18 key points, so The AR game has to be 2D, incompatible with Disco Arena.
Reflected on these constraints, we iterated our design. Our goal is to avoid rotation, and also big movement. Two things we changed in design is, we moved the filter from attached to body to a background, so it won’t be affected by rotation. Second, we moved refillment near the player, to avoid big movement.
Also, based on our previous design, alan came up with our very first prototype to show the core mechanism: