Enhancements error viz

[eleniv3d]

Still work in progress. I open it for discussion and in order to keep track of what's missing

• get rid of o_success for cvs exporter Getting rid of o_success output for csv exporter #86
• add x axis of graph for error visualization + unit Adding the x axis of the graph for the error visualization + unit #85 Missing unit scale in the legend #81
• legend scale show decimals The legend scale is not showing lower decimals #84
• missing o in output of legend Tiny one 🐞 : missing "o" on the output component _legend #80
• Allow the Viz component to show either beams-joints-assembly value graphs Allow the Viz component to show either beams-joints-assembly value graphs #88
  -> implemented as a per_item toggle
• per face metric when analyzing joints per-face metric when analysing the joints #89
  -> implemented if pcd input is already segmented per face

[DamienGilliard]

@eleniv3d Thanks ! I took the liberty of linking the issues, so they are closed automatically when merging.

I was just looking at the per-face metric issue. Do you see a clear way to solve it ? I'm just diving in and it's not 100% clear to me how we can do that the most efficiently.

[eleniv3d]

@eleniv3d Thanks ! I took the liberty of linking the issues, so they are closed automatically when merging.

I was just looking at the per-face metric issue. Do you see a clear way to solve it ? I'm just diving in and it's not 100% clear to me how we can do that the most efficiently.

@DamienGilliard
I guess one way could be to already segment the pcd per face with a similar mechanism as for the joints. Perhaps the joint segmentator could already output clusters for faces directly if it doesn't add to the calculation time. Then the pcd2mesh distance would automatically know that it is calculating for every face and will pop the corresponding message as is now the case for joints vs beams

The other way would be to specify that we want the calculation to happen per face with another toggle or ideally if we manage to make issue #116 work. but then again the distance calculation would look in the whole pcd of the joint.

[9and3]

@DamienGilliard
I guess one way could be to already segment the pcd per face with a similar mechanism as for the joints. Perhaps the joint segmentator could already output clusters for faces directly if it doesn't add to the calculation time. Then the pcd2mesh distance would automatically know that it is calculating for every face and will pop the corresponding message as is now the case for joints vs beams

can we not use the DFFace's brep face instead of the entire brep when doing the calculation of the error @eleniv3d ? And repeat it for each DFFace. I have some doubts about having an extra clustering component just for faces.

The other way would be to specify that we want the calculation to happen per face with another toggle or ideally if we manage to make issue #116 work. but then again the distance calculation would look in the whole pcd of the joint.

This could be nice and easy the same way we did it for the VIzSettings

[DamienGilliard]

@eleniv3d we do agree that you want a list of lists of pointclouds (and None if no pointcloud was found)? :
[[face_1_joint_1, face_2_joint_1, ...], [face_1_joint_2, face_2_joint_2, ...], ..., [face_1_joint_n, face_2_joint_n...]]
(Just double checking, I am the kind of person who messes this kind of things easily)

[eleniv3d]

@eleniv3d we do agree that you want a list of lists of pointclouds (and None if no pointcloud was found)? : [[face_1_joint_1, face_2_joint_1, ...], [face_1_joint_2, face_2_joint_2, ...], ..., [face_1_joint_n, face_2_joint_n...]] (Just double checking, I am the kind of person who messes this kind of things easily)

Sounds good!

[eleniv3d]

Possibly okay to review now. I struggled a bit with getting the desired data from the segmentators and in the end hacked it a bit. Maybe we can see together in the next meeting.

Notes

1. The input of the source point cloud can now be joints, beams or faces of any data structure. We internally flatten it and always compare it with the length of the corresponding lists from the assembly
2. We handle the exception if a pcd is invalid and output the same invalid pcd or color its "twin" mesh white
3. I'll do a separate PR for the CVS exporter

(per face calculation)

(invalid items visualization on pcd)

(invalid items visualization on mesh)

[9and3]

hello @eleniv3d ! Thanks for this PR, I just left some comments and did some tiny cleaning myself! Merging this..

[9and3]

to remove

[9and3]

to remove

[9and3]

In line 22 I propose to add this to avoid drop-on-canvas error:

if i_cloud_source is None or i_assembly is None:
            return None, None, None, None, None, None

[9and3]

to remove

[9and3]

the same we could add:

if i_result or i_viz_settings:
    return None, None