Super4PCS and 4PCS give very different results?

[wenbowen123]

I'm running the "run-example.sh" file under "scripts/"
And here is the output. The transformation given by super4pcs and 4pcs are very different. And 4pcs is visually proved to be correct. What could be the reason for this?

Running Super4PCS
Starting registration
norm_max_dist: 0.01
Initial LCP: 0.04
Score: 0.1best: 0.100000                  
(Homogeneous) Transformation from ../assets/hippo2.obj to ../assets/hippo1.obj: 
-0.753964   0.26322  0.601874  0.161226
 0.110139  0.953895 -0.279201  0.164067
-0.647617 -0.144218 -0.748194 0.0663574
        0         0         0         1
Exporting Matrix to mat_super4pcs_fast.txt...
Export DONE
Exporting Registered geometry to super4pcs_fast.obj...
Export DONE
real 0.44
user 0.33
sys 0.10
------------------------------
Running 4PCS
Starting registration
norm_max_dist: 0.01
Initial LCP: 0.04
Score: 0.465st: 0.465000                  
(Homogeneous) Transformation from ../assets/hippo2.obj to ../assets/hippo1.obj: 
    0.785391     -0.07766    -0.614109   -0.0822443
   0.0600587     0.996978   -0.0492675 -0.000971448
    0.616079   0.00181172     0.787682   -0.0175997
           0            0            0            1
Exporting Matrix to mat_4pcs_fast.txt...
Export DONE
Exporting Registered geometry to 4pcs_fast.obj...
Export DONE
real 0.54
user 0.43
sys 0.10

[nmellado]

Thanks for the report.
Indeed this is not expected.

Could you please post here information regarding your compiler (name, version), compilation option (if any) and compilation mode (debug, release), operating system and OpenGR commit sha ?
Also, which version of Eigen do you use ?

Thanks

[wenbowen123]

@nmellado
gcc 4.9
default compilation option in CMakeLists.
compilation mode: release
Ubuntu 16.04
commit: 5c6560e
Eigen 3

[nmellado]

Ok thanks, I'll look at it asap.