Below some approximate convex decomposition results comparing
V-HACD 2.0 and
HACD.
Parameters:
- testVHACD.exe %%i 8000000 20 0.003 4 4 0.05 0.05 0.0015 0 0 256 0.0 %%i.wrl log_%%i.txt
- testHACD.exe %%i 2 50 0 0 1 30 2000
Score:
- +1: V-HACD provides better decomposition than HACD
- -1: HACD provides better decomposition than V-HACD
- 0: V-HACD and HACD provide comparable results
V-HACD 2.0 |
HACD |
Score |
18 convex-hulls
|
66 convex-hulls
|
1 |
18 convex-hulls
|
26 convex-hulls
|
1 |
16 convex-hulls
|
12 convex-hulls
|
-1 |
30 convex-hulls
|
28 convex-hulls
|
0 |
46 convex-hulls
|
54 convex-hulls
|
1 |
18 convex-hulls
|
26 convex-hulls
|
1 |
16 convex-hulls
|
19 convex-hulls
|
0 |
18 convex-hulls
|
22 convex-hulls
|
0 |
34 convex-hulls
|
17 convex-hulls
|
1 |
18 convex-hulls
|
13 convex-hulls
|
-1 |
25 convex-hulls
|
28 convex-hulls
|
1 |
20 convex-hulls
|
15 convex-hulls
|
-1 |
22 convex-hulls
|
16 convex-hulls
|
-1 |
42 convex-hulls
|
42 convex-hulls
|
0 |
9 convex-hulls
|
7 convex-hulls
|
-1 |
16 convex-hulls
|
15 convex-hulls
|
1 |
35 convex-hulls
|
37 convex-hulls
|
0 |
|
17 convex-hulls
|
17 convex-hulls
|
0 |
|
18 convex-hulls
|
18 convex-hulls
|
0 |
36 convex-hulls
|
34 convex-hulls
|
0 |
13 convex-hulls
|
10 convex-hulls
|
-1 |
11 convex-hulls
|
5 convex-hulls
|
-1 |
24 convex-hulls
|
20 convex-hulls
|
1 |
35 convex-hulls
|
36 convex-hulls
|
0 |
21 convex-hulls
|
13 convex-hulls
|
-1 |
15 convex-hulls
|
16 convex-hulls
|
0 |
9 convex-hulls |
10 convex-hulls
|
0 |
41 convex-hulls
|
71 convex-hulls
|
0 |
22 convex-hulls
|
27 convex-hulls
|
0 |
47 convex-hulls
|
51 convex-hulls
|
0 |
23 convex-hulls
|
28 convex-hulls
|
0 |
Very nice results!
ReplyDeleteHave you changed the Convex hull computer function in V-HACD 2.0?
I am using the convex hull class provided in HACD and was wondering if I need to upgrade to this library.
Hi Nilay,
ReplyDeleteYes I am using the comvex-hull code provided by bullet http://bulletphysics.org/Bullet/BulletFull/btConvexHullComputer_8h.html
I think that V-HACD provides more consistent results. For some meshes, HACD may provide more accurate results for the same number of clusters (see the bunny model above).
Please, give it a try and let me know what you think.
How does the collision detection work between the convex bodies after decomposition ? Do we need to use GJK for this purpose or some other approach
ReplyDelete