Wrong UV texture coordinates when importing VRML file #39596

New Issue

Guillaume Raffy · 2014-04-04T18:10:53+02:00

Guillaume Raffy commented

2014-04-04 18:10:53 +02:00

System Information
MacOSX 10.6.8 on Apple MacBook 6,1

Blender Version
Broken: 2.70 19e627c

Short description of error
The VRML importer provided as an add-on with blender doesn't import texture coordinates properly : the texture coordinates indices are off by 1.

Exact steps for others to reproduce the error
- start blender
- import a VRML scene that contains texture nodes (for example, fig6.vrml that I uploaded with the case #39594 : once the fix to #39594 is applied, we can see in blender that the texture doesn't display properly (the 4 texture coordinates are off by 1), although this scene is displayed properly with the FreeWRL software)

I attached the patch that I made. This patch fixes this bug, but I beleive that there are some other similar bugs remaining, as this bug seems to have been introduced in an attempt to fix bug blender/blender#18329, as it was tagged with the string 'EEKADOODLE'

texcoords_patch_to_v20131007.patch

texcoords_patch_to_v20131007.unifiedpatch

**System Information** MacOSX 10.6.8 on Apple MacBook 6,1 **Blender Version** Broken: 2.70 19e627c **Short description of error** The VRML importer provided as an add-on with blender doesn't import texture coordinates properly : the texture coordinates indices are off by 1. **Exact steps for others to reproduce the error** - start blender - import a VRML scene that contains texture nodes (for example, fig6.vrml that I uploaded with the case #39594 : once the fix to #39594 is applied, we can see in blender that the texture doesn't display properly (the 4 texture coordinates are off by 1), although this scene is displayed properly with the FreeWRL software) I attached the patch that I made. This patch fixes this bug, but I beleive that there are some other similar bugs remaining, as this bug seems to have been introduced in an attempt to fix bug blender/blender#18329, as it was tagged with the string 'EEKADOODLE' [texcoords_patch_to_v20131007.patch](https://archive.blender.org/developer/F83943/texcoords_patch_to_v20131007.patch) [texcoords_patch_to_v20131007.unifiedpatch](https://archive.blender.org/developer/F83944/texcoords_patch_to_v20131007.unifiedpatch)

Guillaume Raffy commented

2014-04-04 18:10:53 +02:00

Changed status to: 'Open'

Guillaume Raffy commented

2014-04-04 18:10:53 +02:00

Added subscriber: @graffy

Thomas Dinges commented

2014-04-08 17:12:28 +02:00

Added subscriber: @ThomasDinges

Jonatan Bijl commented

2014-04-16 14:31:37 +02:00

Added subscriber: @jonim8or

Jonatan Bijl commented

2014-04-16 14:31:37 +02:00

Are you sure the UVs in your file fig6.vrml are correct? because the uvs from the laetetia model http://www.web3d.org/x3d/content/examples/Basic/StudentProjects/_pages/page23.html appear to be imported fine. (and get broken with your version)

Guillaume Raffy commented

2014-04-17 10:18:27 +02:00

To be honest, I'm not 100% percent sure that fig6.vrml has correct UVs (this vrml file has been generated by mayavi, which is based on vtk, so there might well be a bug in their code). The only reason that makes me think that fig6.vrml is correct is that the UVs look fine when loaded into FreeWRL (http://en.wikipedia.org/wiki/FreeWRL), which seems serious enough...

The Laetetia model http://www.web3d.org/x3d/content/examples/Basic/StudentProjects/_pages/page23.html appears fine in FreeWRL, which seems to confirm that this model contains proper vrml.

As a result, it seems that my patch is suboptimal : it breaks laetitia import while it fixes fig6.vrml.

On a side note when made my 'fix', I noticed that a dummy texture coordinates pair was added to the list of textures coordinates in order to fix the 1-based indices issue. This is quite messy and I don't think this is necessary. Unless I overlooked something, this issue can be solved solely by using additions to UV coordinates indices (keep all python lists 0-based, and adapt indices only at the vrml parsing stage if vrml's indices are 1-based)

To be honest, I'm not 100% percent sure that fig6.vrml has correct UVs (this vrml file has been generated by mayavi, which is based on vtk, so there might well be a bug in their code). The only reason that makes me think that fig6.vrml is correct is that the UVs look fine when loaded into FreeWRL (http://en.wikipedia.org/wiki/FreeWRL), which seems serious enough... The Laetetia model http://www.web3d.org/x3d/content/examples/Basic/StudentProjects/_pages/page23.html appears fine in FreeWRL, which seems to confirm that this model contains proper vrml. As a result, it seems that my patch is suboptimal : it breaks laetitia import while it fixes fig6.vrml. On a side note when made my 'fix', I noticed that a dummy texture coordinates pair was added to the list of textures coordinates in order to fix the 1-based indices issue. This is quite messy and I don't think this is necessary. Unless I overlooked something, this issue can be solved solely by using additions to UV coordinates indices (keep all python lists 0-based, and adapt indices only at the vrml parsing stage if vrml's indices are 1-based)

Jonatan Bijl commented

2014-04-17 10:58:40 +02:00

I really wonder if the tex coords are 1-based. From the vrml specification I understand that they are 0-based:

 If the greatest index in the coordIndex field is N, then the Coordinate node must contain N+1 coordinates (indexed as 0-N)

That should also go for the vertex colors, according to the specification. But I guess there's a reason why the eekadoodle was added.

In my version of the importer #38261 I have removed the dummy entry for uvs, and the "+ 1" that was done when referring to them. And it seems to work (with the laetitia model).

I really wonder if the tex coords are 1-based. From [the vrml specification](http://graphcomp.com/info/specs/sgi/vrml/spec/part1/nodesRef.html#IndexedFaceSet) I understand that they are 0-based: ``` If the greatest index in the coordIndex field is N, then the Coordinate node must contain N+1 coordinates (indexed as 0-N) ``` That should also go for the vertex colors, according to the specification. But I guess there's a reason why the eekadoodle was added. In my version of the importer #38261 I have removed the dummy entry for uvs, and the "+ 1" that was done when referring to them. And it seems to work (with the laetitia model).

Brendon Murphy commented

2016-01-08 07:46:08 +01:00

Added subscriber: @BrendonMurphy

Brendon Murphy commented

2016-01-08 07:46:08 +01:00

Changed status from 'Open' to: 'Archived'

Brendon Murphy closed this issue

2016-01-08 07:46:08 +01:00

Brendon Murphy self-assigned this 2016-01-08 07:46:08 +01:00

Brendon Murphy commented

2016-01-08 07:46:08 +01:00

closing as archived, no activity 1.5 years

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Wrong UV texture coordinates when importing VRML file #39596