Can't get proper result from the test data

[az508]

Hi:

I've tried the new version with the test data, but can't get correct result. The reprojection error is extremely large (about 8000 pixel) and the estimated parameters looks strange

`R =

-0.2894 0.3261 0.8999 -0.7342 -0.6789 0.0099 0.6142 -0.6578 0.4359

T =

1.0e+132 *

-0.3988 -0.0349 -1.2250

n1 =

-0.2983 -0.0369 -0.9538

n2 =

-0.3227 -0.0536 -0.9450

n3 =

-0.3268 -0.0674 -0.9427

d1 =

-6.4316e+131

d2 =

-6.4322e+131

d3 =

-6.4308e+131`

Matlab showed some message like `Warning: Matrix is close to singular or badly scaled. Results may be inaccurate. RCOND = 2.042577e-26.

In gauss_newton at 34 In optimize_betas_gauss_newton at 29 In efficient_pnp_gauss at 216 In tnm at 60 In demo at 53 `

Do you have any idea?

Best

[qtarou]

Hi. Thank you for the comment.

I'm sorry. I have fixed the bug. Please get updated code and try it. (I replaced the efficient_pnp_gauss() function to efficient_pnp() in tnm.m. Notice that the estimation of 3D point (i.e. EPnP) are not our contributions and you can replace this function with other algorithm for more robust estimation.)

Best

[az508]

Thank you for the reply. I get correct result now.

Best