Alembic not importing UVs (if file incorrectly sets geoScope to kVertexScope - even though actual data implies kFacevaryingScope - which 2.93 handled gracefully) #93906

New Issue

zhangliqi · 2021-12-09T16:25:59+01:00

zhangliqi commented

2021-12-09 16:25:59 +01:00

System Information
Operating system: Windows-10-10.0.19041-SP0 64 Bits
Graphics card: NVIDIA GeForce RTX 3070 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 472.19

Blender Version
Broken: version: 3.0.0, branch: master, commit date: 2021-12-02 18:35, hash: f1cca30557
Worked: 2.93

Since 3385c04598 we are more strict in regards to "malformed" .abc (see discussion below)

Short description of error
in 2.93 import no question
but 3.0 uv data lost
abc file{F12721151}

**System Information** Operating system: Windows-10-10.0.19041-SP0 64 Bits Graphics card: NVIDIA GeForce RTX 3070 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 472.19 **Blender Version** Broken: version: 3.0.0, branch: master, commit date: 2021-12-02 18:35, hash: `f1cca30557` Worked: 2.93 Since 3385c04598 we are more strict in regards to "malformed" .abc (see discussion below) **Short description of error** in 2.93 import no question but 3.0 uv data lost abc file{[F12721151](https://archive.blender.org/developer/F12721151/alembic_file.abc)}

zhangliqi commented

2021-12-09 16:25:59 +01:00

Added subscriber: @shadow_creeper

zhangliqi changed title from ~~abc import mmd 同~~ to import mmd turn to abc ，uv data lost

2021-12-09 16:43:34 +01:00

Richard Antalik commented

2021-12-09 22:25:37 +01:00

Changed status from 'Needs Triage' to: 'Confirmed'

Philipp Oeser commented

2021-12-10 09:25:29 +01:00

Added subscriber: @lichtwerk

Philipp Oeser self-assigned this 2021-12-10 09:25:29 +01:00

Philipp Oeser commented

2021-12-10 09:25:29 +01:00

Will check.

Philipp Oeser commented

2021-12-10 11:08:53 +01:00

So, first findings:

importing this .abc was fine in 2.93, then
- started crashing with f9567f6c63
- crash was fixed by 7c68147709 (but no more UVs)

If I compare a simple .abc export of a cube with UVs [current master] and provided .abc I see the following:


/Cube/Cube/.geom:
ScalarProperty      float64_t[6]   .selfBnds[1] {interpretation=box}
ArrayProperty       float32_t[3]   P[1] {
                                    geoScope=vtx
                                    interpretation=point
                                   }
ArrayProperty       int32_t        .faceIndices[1]
ArrayProperty       int32_t        .faceCounts[1]
CompoundProperty                   .userProperties
CompoundProperty                   .arbGeomParams
CompoundProperty                   uv {
                                    arrayExtent=1
                                    geoScope=fvr
                                    interpretation=vector
                                    isGeomParam=true
                                    podExtent=2
                                    podName=float32_t
                                    sourceName=UVMap
                                   }
ArrayProperty       float32_t[3]   N[1] {
                                    arrayExtent=1
                                    geoScope=fvr
                                    interpretation=normal
                                    isGeomParam=true
                                    podExtent=3
                                    podName=float32_t
                                   }
/Cube/Cube/.userProperties:
ScalarProperty      bool_t         meshtype[1]
/Cube/Cube/uv:
ArrayProperty       float32_t[2]   .vals[1] {
                                    arrayExtent=1
                                    geoScope=fvr
                                    interpretation=vector
                                    isGeomParam=true
                                    podExtent=2
                                    podName=float32_t
                                    sourceName=UVMap
                                   }

/xform_0_material_0/mesh_0_material_0/.geom:
ScalarProperty      float64_t[6]   .selfBnds[101] {interpretation=box}
ArrayProperty       float32_t[3]   P[101] {
                                    geoScope=vtx
                                    interpretation=point
                                   }
ArrayProperty       int32_t        .faceIndices[101]
ArrayProperty       int32_t        .faceCounts[101]
ArrayProperty       float32_t[2]   uv[101] {
                                    arrayExtent=1
                                    geoScope=vtx
                                    interpretation=vector
                                    isGeomParam=true
                                    podExtent=2
                                    podName=float32_t
                                   }
ArrayProperty       float32_t[3]   N[101] {
                                    arrayExtent=1
                                    geoScope=vtx
                                    interpretation=normal
                                    isGeomParam=true
                                    podExtent=3
                                    podName=float32_t
                                   }

Seems like UVs are stored in plain ArrayProperty (instead of CompoundProperty / userProperties), also the sourceName is missing.
Checking why this was still importing fine before f9567f6c63 now.

@shadow_creeper : was this .abc exported from blender? (which version?)

So, first findings: - importing this .abc was fine in 2.93, then - started crashing with f9567f6c63 - crash was fixed by 7c68147709 (but no more UVs) If I compare a simple .abc export of a cube with UVs [current master] and provided .abc I see the following: ``` /Cube/Cube/.geom: ScalarProperty float64_t[6] .selfBnds[1] {interpretation=box} ArrayProperty float32_t[3] P[1] { geoScope=vtx interpretation=point } ArrayProperty int32_t .faceIndices[1] ArrayProperty int32_t .faceCounts[1] CompoundProperty .userProperties CompoundProperty .arbGeomParams CompoundProperty uv { arrayExtent=1 geoScope=fvr interpretation=vector isGeomParam=true podExtent=2 podName=float32_t sourceName=UVMap } ArrayProperty float32_t[3] N[1] { arrayExtent=1 geoScope=fvr interpretation=normal isGeomParam=true podExtent=3 podName=float32_t } /Cube/Cube/.userProperties: ScalarProperty bool_t meshtype[1] /Cube/Cube/uv: ArrayProperty float32_t[2] .vals[1] { arrayExtent=1 geoScope=fvr interpretation=vector isGeomParam=true podExtent=2 podName=float32_t sourceName=UVMap } ``` ``` /xform_0_material_0/mesh_0_material_0/.geom: ScalarProperty float64_t[6] .selfBnds[101] {interpretation=box} ArrayProperty float32_t[3] P[101] { geoScope=vtx interpretation=point } ArrayProperty int32_t .faceIndices[101] ArrayProperty int32_t .faceCounts[101] ArrayProperty float32_t[2] uv[101] { arrayExtent=1 geoScope=vtx interpretation=vector isGeomParam=true podExtent=2 podName=float32_t } ArrayProperty float32_t[3] N[101] { arrayExtent=1 geoScope=vtx interpretation=normal isGeomParam=true podExtent=3 podName=float32_t } ``` Seems like UVs are stored in plain `ArrayProperty` (instead of `CompoundProperty` / `userProperties`), also the `sourceName` is missing. Checking why this was still importing fine before f9567f6c63 now. @shadow_creeper : was this .abc exported from blender? (which version?)

zhangliqi commented

2021-12-10 11:17:39 +01:00

i use mikumikudance softwave of mikumiku bridge addon , .mmd exported to .abc
2.93 or c4d can import this abc
but 3.0 import will lost uv data
@philipp-14 Oeser (lichtwerk)

i use mikumikudance softwave of mikumiku bridge addon , .mmd exported to .abc 2.93 or c4d can import this abc but 3.0 import will lost uv data @philipp-14 Oeser (lichtwerk)

zhangliqi commented

2021-12-10 11:18:30 +01:00

Added subscriber: @philipp-14

zhangliqi commented

2021-12-10 11:24:59 +01:00

@philipp-14 Oeser
is it pssoble to do a test and use 2.93 code ti import if uv data is lost?

@philipp-14 Oeser is it pssoble to do a test and use 2.93 code ti import if uv data is lost?

Philipp Oeser commented

2021-12-10 11:31:53 +01:00

@shadow_creeper : I am on it, please use "lichtwerk" as my user name (poor Philipp Kant has been subscribed on accident to too many reports, sorry for this)

As it looks now, the real culprit commit is 3385c04598, but checking further...

@shadow_creeper : I am on it, please use "lichtwerk" as my user name (poor Philipp Kant has been subscribed on accident to too many reports, sorry for this) As it looks now, the real culprit commit is 3385c04598, but checking further...

Philipp Oeser commented

2021-12-10 12:27:55 +01:00

Changed status from 'Confirmed' to: 'Needs User Info'

Philipp Oeser commented

2021-12-10 12:27:55 +01:00

Added subscriber: @dr.sybren

Philipp Oeser commented

2021-12-10 12:27:55 +01:00

So the real "problem" is actually in the file I think.
It defines the geoScope of the ArrayProperty as being kVertexScope:

geoScope=vtx

The indices look like they are meant for face corners though (they dont match the vertex count, but they match the loop count).
Prior to 3385c04598, the scope wasnt checked as thoroughly, so face corners were always assumed [ and thus nobody noticed ;) -- everything was just read in as facecorner/loop data and all was good].
However, this is still an error in the file, if it exports the UVs on face corners, the scope should be kFacevaryingScope [ reg. C4D -- i can only assume they are sloopy as we were before 3385c04598 :) ]:

geoScope=fvr

I can get it to import just fine doing the following hack [@dr.sybren: this will never get allowed in master, right?]
P2656: T93906_hack



diff --git a/source/blender/io/alembic/intern/abc_customdata.cc b/source/blender/io/alembic/intern/abc_customdata.cc
index 830ec731e20..5d381a9aee3 100644
--- a/source/blender/io/alembic/intern/abc_customdata.cc
+++ b/source/blender/io/alembic/intern/abc_customdata.cc
@@ -611,6 +611,11 @@ AbcUvScope get_uv_scope(const Alembic::AbcGeom::GeometryScope scope,
     return ABC_UV_SCOPE_LOOP;
   }
 
+  /* Some software seems to be sloppy [writes loop UVs but specifies vertex scope...]. */
+  if (scope == kVertexScope && indices->size() == config.totloop) {
+    return ABC_UV_SCOPE_LOOP;
+  }
+
   /* kVaryingScope is sometimes used for vertex scopes as the values vary across the vertices. To
    * be sure, one has to check the size of the data against the number of vertices, as it could
    * also be a varying attribute across the faces (i.e. one value per face). */

Anyways, to understand the mikumiku workflow better:

what is actually doing the .abc export? blender? which blender version?
- if so, please provide an example .blend file (where the MDD is already imported), the MDD -- just prior to exporting the abc
- also provide the abc export settings then
if not: which software produces the .abc?

So the real "problem" is actually in the file I think. It defines the `geoScope` of the `ArrayProperty` as being `kVertexScope`: ``` geoScope=vtx ``` The indices look like they are meant for face corners though (they dont match the vertex count, but they match the loop count). Prior to 3385c04598, the scope wasnt checked as thoroughly, so face corners were always assumed [ and thus nobody noticed ;) -- everything was just read in as facecorner/loop data and all was good]. However, this is still an error in the file, if it exports the UVs on face corners, the scope should be `kFacevaryingScope` [ reg. C4D -- i can only assume they are sloopy as we were before 3385c04598 :) ]: ``` geoScope=fvr ``` I can get it to import just fine doing the following hack [@dr.sybren: this will never get allowed in master, right?] [P2656: T93906_hack](https://archive.blender.org/developer/P2656.txt) ``` diff --git a/source/blender/io/alembic/intern/abc_customdata.cc b/source/blender/io/alembic/intern/abc_customdata.cc index 830ec731e20..5d381a9aee3 100644 --- a/source/blender/io/alembic/intern/abc_customdata.cc +++ b/source/blender/io/alembic/intern/abc_customdata.cc @@ -611,6 +611,11 @@ AbcUvScope get_uv_scope(const Alembic::AbcGeom::GeometryScope scope, return ABC_UV_SCOPE_LOOP; } + /* Some software seems to be sloppy [writes loop UVs but specifies vertex scope...]. */ + if (scope == kVertexScope && indices->size() == config.totloop) { + return ABC_UV_SCOPE_LOOP; + } + /* kVaryingScope is sometimes used for vertex scopes as the values vary across the vertices. To * be sure, one has to check the size of the data against the number of vertices, as it could * also be a varying attribute across the faces (i.e. one value per face). */ ``` Anyways, to understand the mikumiku workflow better: - what is actually doing the .abc export? blender? which blender version? - if so, please provide an example .blend file (where the MDD is already imported), the MDD -- just prior to exporting the abc - also provide the abc export settings then - if not: which software produces the .abc?

Philipp Oeser removed their assignment 2021-12-17 10:47:55 +01:00

Philipp Oeser commented

2021-12-17 10:47:55 +01:00

Still waiting for answers to this:

In #93906#1270093, @lichtwerk wrote:
Anyways, to understand the mikumiku workflow better:

what is actually doing the .abc export? blender? which blender version?
-- if so, please provide an example .blend file (where the MDD is already imported), the MDD -- just prior to exporting the abc
-- also provide the abc export settings then

if not: which software produces the .abc?

For until then, will step down as the one assigned here...

Still waiting for answers to this: > In #93906#1270093, @lichtwerk wrote: > Anyways, to understand the mikumiku workflow better: > - what is actually doing the .abc export? blender? which blender version? > -- if so, please provide an example .blend file (where the MDD is already imported), the MDD -- just prior to exporting the abc > -- also provide the abc export settings then > - if not: which software produces the .abc? For until then, will step down as the one assigned here...

Philipp Oeser changed title from ~~import mmd turn to abc ，uv data lost~~ to Alembic not importing UVs (if file incorrectly sets geoScope to kVertexScope - even though actual data implies kFacevaryingScope - which 2.93 handled gracefully)

2021-12-17 10:49:43 +01:00

Philipp Oeser commented

2022-01-10 10:20:31 +01:00

Changed status from 'Needs User Info' to: 'Archived'

Philipp Oeser closed this issue

2022-01-10 10:20:31 +01:00

Philipp Oeser commented

2022-01-10 10:20:31 +01:00

No activity for more than a week. As per the tracker policy we assume the issue is gone and can be closed.

Thanks again for the report. If the problem persists please open a new report with the required information.

No activity for more than a week. As per the tracker policy we assume the issue is gone and can be closed. Thanks again for the report. If the problem persists please open a new report with the required information.

Kotori commented

2023-05-26 00:05:56 +02:00

Still waiting for answers to this:

In #93906#1270093, @lichtwerk wrote:
Anyways, to understand the mikumiku workflow better:

what is actually doing the .abc export? blender? which blender version?
-- if so, please provide an example .blend file (where the MDD is already imported), the MDD -- just prior to exporting the abc
-- also provide the abc export settings then

if not: which software produces the .abc?

For until then, will step down as the one assigned here...

I can provide you with some answers.
https://github.com/uimac/mmdbridge
In the link above is the source code of the software used in the report, where you can find some relevant information.
It is a software accessory plug-in, you can refer to the link below for the operation (but it does not help to solve the problem).
https://mmdbridge.vec4d.xyz/%E3%83%9B%E3%83%BC%E3%83%A0/usage-in-english
I need to use this software to export .abc because of some project requirements. Unfortunately, it still loses uv in Blender3.5 version (it is normal in 2.93 and other 3D software).
As you said, it might be a problem in the abc file. But it's not necessarily a bug, it could be to simplify the file size. And if it's a bug, you should get the same result in other 3D software.

Refer to the above script screenshot and source code file (you need to open the project file with VS to view the reference of d3d9), we only study mode 0 and 2, the behavior of mode 1 is very strange, and the result of mode 3 is the same as that of 0, so we don’t consider it for now. The case of mode 0 may be related to the loop count you said (it will be represented as missing uv in 3.5), and the mode 2 may be related to the vertex count (it works normally in 3.5). But the former only needs 200mb to export the .abc file of 100 frames, and the latter needs nearly 1000mb. This is a challenge for my hard drive.

Finally, you can find information about kVertexScope in the screenshot above (it's located in d3d9/alembic.cpp), but since I'm not a professional, I can't understand it.
I think it is necessary to reopen the report, hope this reply helps you to fix the error.

> Still waiting for answers to this: > > > In #93906#1270093, @lichtwerk wrote: > > Anyways, to understand the mikumiku workflow better: > > - what is actually doing the .abc export? blender? which blender version? > > -- if so, please provide an example .blend file (where the MDD is already imported), the MDD -- just prior to exporting the abc > > -- also provide the abc export settings then > > - if not: which software produces the .abc? > > > For until then, will step down as the one assigned here... I can provide you with some answers. https://github.com/uimac/mmdbridge In the link above is the source code of the software used in the report, where you can find some relevant information. It is a software accessory plug-in, you can refer to the link below for the operation (but it does not help to solve the problem). https://mmdbridge.vec4d.xyz/%E3%83%9B%E3%83%BC%E3%83%A0/usage-in-english I need to use this software to export .abc because of some project requirements. Unfortunately, it still loses uv in Blender3.5 version (it is normal in 2.93 and other 3D software). As you said, it might be a problem in the abc file. But it's not necessarily a bug, it could be to simplify the file size. And if it's a bug, you should get the same result in other 3D software. ![1.png](/attachments/d6a5c57e-a897-4658-b1cc-7d00f52e836f) ![2.png](/attachments/d8948e29-2ca5-4aa8-af62-093fa477cee6) Refer to the above script screenshot and source code file (you need to open the project file with VS to view the reference of d3d9), we only study mode 0 and 2, the behavior of mode 1 is very strange, and the result of mode 3 is the same as that of 0, so we don’t consider it for now. The case of mode 0 may be related to the loop count you said (it will be represented as missing uv in 3.5), and the mode 2 may be related to the vertex count (it works normally in 3.5). But the former only needs 200mb to export the .abc file of 100 frames, and the latter needs nearly 1000mb. This is a challenge for my hard drive. ![3.png](/attachments/7fc3ec01-7a92-4585-b84c-5fd4489f50aa) Finally, you can find information about kVertexScope in the screenshot above (it's located in d3d9/alembic.cpp), but since I'm not a professional, I can't understand it. I think it is necessary to reopen the report, hope this reply helps you to fix the error.

1.png

45 KiB

3.png

73 KiB

2.png

61 KiB

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

Alembic not importing UVs (if file incorrectly sets geoScope to kVertexScope - even though actual data implies kFacevaryingScope - which 2.93 handled gracefully) #93906