yqueau / near_ps Goto Github PK
View Code? Open in Web Editor NEWCode for solving photometric stereo under calibrated or semi-calibrated near point light source illumination (e.g., LEDs).
License: GNU General Public License v3.0
Code for solving photometric stereo under calibrated or semi-calibrated near point light source illumination (e.g., LEDs).
License: GNU General Public License v3.0
I found that only qualitative results are provided in the implementation, such as the output obj file. However, is there any way to test it quantitively?
Actually, I try a synthetic sphere case under near-light setting, the normal angular error is 3-4 degree, is that correct?
Hello Yvain,
First of all, thanks for sharing your code.
Nice implementation.
I was playing with your dataset Statuette
I realized a problem when using the cmg preconditioner.
"The current version of CMG does not support positive off-diagonals"
When this happens, the value for precond_L = cmg_sdd(M) is an empty matrix.
This is because a flag is raised inside the cmg_precondition function.
Then the code continues and a solution is reached but I guess the preconditioner is not doing his job.
Did you realized this? Is there any way to fix this issue in order to use the preconditioner?
Since I already opened the issue, I use the occasion to ask you:
I want to use your code for a pavement reconstruction task at a short distance (~200mm).
So the surface lies on a flat plane.
Do you have any advice for the task?
What parameters values (lambda, z0, tolerance, number of iterations,estimator and preconditionre) do you recomend to use for high quality results?
Thanks in advance!
Hello,
I want to use this code with my own dataset. I use that ring led to generate images.
He is the ring I use:
https://www.adafruit.com/product/2874
I setup my camera in the center of that ring. I try to fill manually the calibrated parameter. Here is the list of parameters you mention:
`Calibrated LEDs' parameters are provided in the file light.mat, which contains the following variables:
S (8 x 3): XYZ location of the 8 sources, in mm and w.r.t. the camera center (REQUIRED).
Phi (8 x 3): RGB intensities of the 8 sources, w.r.t. the color of the calibration target we used (assumed to have albedo equal to 1 in each wavelength). If these intensities were not calibrated, they can be estimated automatically using the semi-calibrated option.
mu (8 x 1): anisotropy parameter of the 8 sources. mu(i) = 0 means the it-th source is isotropic, mu =1 means it is a primary Lambertian source, and mu > 1 means it is anisotropic (imperfect Lambertian source model). Set mu(:) = 0 if you have no idea about this parameter.
Dir (8 x 3): XYZ orientation of the 8 sources, w.r.t. the optical axis. Each row must have unit-length. Required if mu > 0, not required if mu=0.`
If I understand properly:
Dir should be equal to [0,0,1] in my setup, S should be the position of the led according to the camera center.
What should be the value Phi and mu ?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.