In each layer of the image pyramid using K-means ?

[payne11]

Besides , i feel not match in this part of the program and the paper. trunc_res = max(0.007f, min(0.015f*(1.f + 10.f*twist_odometry.norm()), median_res)); lim_nobackg = (1.f + 10.f*twist_odometry.norm())*trunc_res; lim_backg = (2.f + 10.f*twist_odometry.norm())*trunc_res according to your program that it should write as follows: δ̂ t = max(t M , min(t B , δ̂)) , δ L = δ̂ t(1+ 10 kξ R k 2) ,δ H = δ̂ t (2 + 10 kξ R k 2) . rather than written in your paper as follows: δ̂ t = min(t M , max(t B , δ̂)) , δ L = δ̂ t + 10 kξ R k 2 ,δ H = 2 δ̂ t + 10 kξ R k 2 . Maybe I have not understood yet, can you explain it to me? Thank you!

[MarianoJT88]

Those thresholds / limits are very heuristic and there might be small differences between the equation presented in the paper and my last implementation on the code. I would keep the limits as they are on the code but you can change and play with those if you want, maybe you get something even better than that.