sigma,relative error and roughness for comparision model

[abhigoku99]

HI,
i am new to the cloud compare .
1.i wanted to know how the sigma value is calculated when i compare two 3D models
2. how is relative error calculated and wat does 3.77496 + 20.1735/d % (d>0.201735) this 'd' represent
2.how can i compare two models based on the roughness and wat is the blue and green color represent in the model

[daniel]

Hi

1. The 'sigma' value is also called the standard deviation (http://en.wikipedia.org/wiki/Standard_deviation). It is computed from all the distances values (one for each point/vertex of the compared entity)
2. Relative error is only meaningful for the approximate distances (quickly computed when you start the tool so as to help CloudCompare guess the best parameters). You should ignore those values.
3. Not sure what you want to do? Can you explain a bit more? Otherwise, to know what colors correspond to, just display the color scale attached to the active scalar field: you have to close the comparison tool, select only the compared entity, and then check the 'Visible' checkbox of the 'Color scale' section of the entity properties (or use the shortcut SHIFT+C).

[abhigoku99]

Thanks for the reply Daniel , below i will mention completely of my agenda of using cloud compare

I have generated a 3D model from my code and i am now comparing it with ground truth data , so i am using cloud compare , when i compare them i will get these values for mean,sigma, max dis,min dis , but i am not able to conclude how good is my model with ground truth based on these values

i have up scaled,down scaled and segmented my model and compared with ground truth and obtained values for all the parameters but looking at that i am not able to evaluate my model can suggest me any idea from these values how can i evaluate my model status compared to ground truth

[daniel]

Have you hit the red 'Compute' button?

Do you see the colored points/vertices/triangles?

Once you get that, you can display the color scale, or an histogram. If ypu don't have enough vertices you can sample points on your mesh.

You can display labels on particular points, and also export the data as txt or csv (to open it in excel or matlab for instance).

[abhigoku99]

HI daniel ,

Yes i have hit the compute button
i see the colored points , i am using histogram also but not able to come into conclusion based on the histogram

i am having two model generated from different codes and i am comparing it with ground truth of the same object , i am registering them to fit into same scale and i am generating the metrics (max,min,avg,sigma,mean) for the two models , for which i am getting different values of metrics for both models how can i say which model is good based on the sigma value

i have attached the excel sheet of the values obtained from cloud compare , how can i concluded which model is better based on the metrics obtained

[daniel]

I don't think there's a 'standard' way to compare errors. It all depends on your 'quality' crietria. And those criteria should depend on what you want to do with the models.

Personally I generally consider the variance (sigma) which gives an idea of the error distribution (if the distribution is Normal) and the maximum error.

In your case the the first model seems much better.

[abhigoku99]

Thank you for the reply