significant overlap among primitives

[qk98]

In the generated result, there is a significant overlap among primitives, such as the chair-leg and back shown in the image. The green primitive appears to be redundant. Is this expected, and how can it be adjusted?

Input:

result:

without GREEN primitive:

without CYAN:

[bmlklwx]

@qk98 Hi, thanks for your interest in our algorithm. This might happen, but could be mitigated by the following tips:
First, allow more iterations for each superquadric optimization (maxOptiIter), which can give each superquadric more chance to converge.
Second, a lower input SDF resolution might help.
Moreover, increase minArea can make the algorithm ignore tiny unfitted volumes where new superquadrics might appear.
Lastly, of course, a post-process will help. For example, check if the removal of a superquadric will affect the overall fitting quality (difference between input and superquadric SDFs). If not, we can remove it. I haven't had time to implement any post-process, but I believe this will help.