Geometry nodes: face corner to point domain iterpolation bug #119986

New Issue

Closed

opened 2024-03-27 22:49:38 +01:00 by baoyu · 10 comments

baoyu commented

2024-03-27 22:49:38 +01:00

Blender Version
Broken: version: 4.2.0 Alpha, branch: main, commit date: 2024-03-27 00:11, hash: ee38a7b723f8

Short description of error
the iterpolation result of face corner domain attribute -> point domain is currently computed using
getPointAttribute( lerp( face_corner.coordinate) )
instead of
lerp( getFaceCornerAttribute( face_corner.coordinate) )

Exact steps for others to reproduce the error
I've made a simple demo:
Using a image to displace the default cylinder, the bug is found at the UV seam.
Ideally the displace hight at the seam should be evaluated like
lerp( getFaceCornerAttribute( [0,0 ]), getFaceCornerAttribute( [0,1 ]) ) which is 0,

but in current GN system, it's evaluated like
getPointAttribute( lerp( [0,0], [0,1] ) ) which is 1.

This behavior is quite counter-intuitive for a typical image based displacement workflow.

**Blender Version** Broken: version: 4.2.0 Alpha, branch: main, commit date: 2024-03-27 00:11, hash: `ee38a7b723f8` **Short description of error** the iterpolation result of face corner domain attribute -> point domain is currently computed using `getPointAttribute( lerp( face_corner.coordinate) )` instead of `lerp( getFaceCornerAttribute( face_corner.coordinate) )` **Exact steps for others to reproduce the error** I've made a simple demo: Using a image to displace the default cylinder, the bug is found at the UV seam. Ideally the displace hight at the seam should be evaluated like `lerp( getFaceCornerAttribute( [0,0 ]), getFaceCornerAttribute( [0,1 ]) )` which is 0, but in current GN system, it's evaluated like `getPointAttribute( lerp( [0,0], [0,1] ) )` which is 1. This behavior is quite counter-intuitive for a typical image based displacement workflow. ![image](/attachments/d3a9be12-0d04-4062-8b50-7a84b784931c)

1007 KiB

face corner to point domain iterpolation bug of set position node.blend

138 KiB

baoyu added the

labels 2024-03-27 22:49:39 +01:00

YimingWu commented

2024-03-28 04:02:27 +01:00

Member

Can confirm. @HooglyBoogly might want to take a look?

Can confirm. @HooglyBoogly might want to take a look?

YimingWu added

Nodes & Physics

and removed

labels 2024-03-28 04:02:44 +01:00

Iliya Katushenock commented

2024-03-28 07:50:03 +01:00

Member

This behavior is quite counter-intuitive for a typical image based displacement workflow.

But this is how field are works. Thanks for report, but this is not a bug.

> This behavior is quite counter-intuitive for a typical image based displacement workflow. But this is how field are works. Thanks for report, but this is not a bug.

Iliya Katushenock closed this issue

2024-03-28 07:50:07 +01:00

Blender Bot added

and removed

labels 2024-03-28 07:50:09 +01:00

baoyu commented

2024-03-28 07:56:32 +01:00

Author

I mean is it really reasonable to lerp the face corner coordinates instead of lerping the attribute itself?

I mean is it really reasonable to lerp the face corner coordinates instead of lerping the attribute itself?

Iliya Katushenock commented

2024-03-28 08:03:45 +01:00

Member

Yes, you read all attributes with interpolation at the begin. And all operations happen only in the same domain for all inputs.

Yes, you read all attributes with interpolation at the begin. And all operations happen only in the same domain for all inputs.

baoyu commented

2024-03-28 08:10:19 +01:00

Author

All right. Then this behavior may cause confusion that it differs from the exisiting displace modifier which actually interpolates like I described above. And it's been used like ages.

All right. Then this behavior may cause confusion that it differs from the exisiting displace modifier which actually interpolates like I described above. And it's been used like ages. ![image](/attachments/331c9db2-10a0-40bf-ad61-e8ad34bd7aab)

959 KiB

Iliya Katushenock commented

2024-03-28 08:12:02 +01:00

Member

This is not expected to be the same.

This is not expected to be the same.

Hans Goudey commented

2024-03-28 20:15:45 +01:00

Member

If you use the "Evaluate on Domain" node to evaluate the image on the corner domain, then do the mixing, that should give the results you expect.

If you use the "Evaluate on Domain" node to evaluate the image on the corner domain, then do the mixing, that should give the results you expect.

baoyu commented

2024-03-29 00:10:21 +01:00

Author

Hi, Hans. All I came up so far is to mix like this.

Can you make a demo file of your method?

If you use the "Evaluate on Domain" node to evaluate the image on the corner domain, then do the mixing, that should give the results you expect.

Hi, Hans. All I came up so far is to mix like this. ![image](/attachments/9e9849dc-99e9-4f95-ba35-799b897a59e0) Can you make a demo file of your method? > If you use the "Evaluate on Domain" node to evaluate the image on the corner domain, then do the mixing, that should give the results you expect.

1.2 MiB

Hans Goudey commented

2024-04-03 15:37:10 +02:00

Member

Sure, this does it:

Sure, this does it: ![image](/attachments/be4434ae-f693-4d8f-85ee-a7caebffffce)

99 KiB

baoyu commented

2024-04-04 06:58:42 +02:00

Author

Ah, how can I missed the explicit evaluate on domain way! Thanks, Hans!

Ah, how can I missed the explicit evaluate on domain way! Thanks, Hans!

Sign in to join this conversation.

No Label

Animation & Rigging

Asset Browser Project Overview

Automated Testing

Blender Asset Bundle

Dependency Graph

Development Management

EEVEE & Viewport

Images & Movies

Motion Tracking

Nodes & Physics

Pipeline, Assets & IO

Platforms, Builds & Tests

Render & Cycles

Render Pipeline

Sculpt, Paint & Texture

Video Sequencer

Virtual Reality

Blender 2.8 Project

Milestone 1: Basic, Local Asset Browser

Good First Issue

Animation & Rigging

Development Management

EEVEE & Viewport

Nodes & Physics

Pipeline, Assets & IO

Platforms, Builds & Tests

Render & Cycles

Sculpt, Paint & Texture

Needs Info from Developers

Needs Information from User

No Milestone

No project

No Assignees

4 Participants

Notifications

Due Date

The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: blender/blender#119986

No description provided.