Question about the DepthCamera.plane2xyz

[bravelywangexocr]

Hi Ankush,

I am wondering the formula in DepthCamera.plane2xyz() for computing "z" by using x, y points and the plane coeffs. The computation is as below:

z = -plane[3]/(xy_ray.dot(plane[:2])+plane[2])

My understanding is that following the plane formula (ax + by + cz = d), the placement of plane[3] (coeff 'd') and plane[2] (coeff 'c') should be exchanged, that means,

z = -plane[2]/(xy_ray.dot(plane[:2])+plane[3])

I did experiments for both cases and plotted the result masks applied by this formula, it appears that sometimes the former works but sometimes the latter does, so I am confused with this.

Can you help confirm this and elaborate more? Or any useful docs for this, please help share.

Thanks.

[ankush-me]

@bravelywangexocr Thank you, you are correct. I think this is a bug, and the code should be updated as you have pointed out. The coefficients [a,b,c,d] correspond to:
ax+ by + cz + d = 0
(you can check here: https://github.com/ankush-me/SynthText/blob/master/ransac.py#L14 )

Did you plot the "depth-image" so obtained?
How did you evaluate which one works better?