Viewer: Align shadow map light with IBL dominant direction #16735
Conversation
|
Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s). |
|
Snapshot stored with reference name: Test environment: To test a playground add it to the URL, for example: https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/16735/merge/index.html#WGZLGJ#4600 Links to test babylon tools with this snapshot: https://playground.babylonjs.com/?snapshot=refs/pull/16735/merge To test the snapshot in the playground with a playground ID add it after the snapshot query string: https://playground.babylonjs.com/?snapshot=refs/pull/16735/merge#BCU1XR#0 |
|
Visualization tests for WebGPU |
|
WebGL2 visualization test reporter: |
|
WebGL2 visualization test reporter: |
|
Visualization tests for WebGPU |
… generator task (BabylonJS#16734) Improved fix for BabylonJS#16710: * The depth texture to use for min/max reduction is now an input of the CSM task. This means that this depth texture can be shared by multiple tasks. In the previous PR, a new depth texture was always generated internally when `autoCalcDepthBounds` was **true** * The textures used for min/max reduction are created as part of the frame graph, which means that these textures can be optimized and reused by the graph I have also added support for “normalized view depth” as an output of the geometry renderer task, as this is the most optimal type for min/max depth reduction when used in CSM.
The previous light direction calculation didn't account for when the Babylon light direction matched glTF's. This would result NaN values in RotationAxisToRef, as the rotation axis would be (0,0,0), and its length (0) used as a divisor. I noticed we already have a function to do what we need for the light direction calculation-- `FromUnitVectorsToRef`-- so I've replaced the whole thing with that. Much simpler. (Bonus: I also noticed that the lights in our glTF light export test don't use glTF falloff, so I've earmarked that change in this PR.)
This PR introduces a test app for local development of Inspector v2. It uses webpack, but a newer version than our other tools which is needed for compatibility with the webpack React fast refresh plugin (which is a new local dev only dependency). The simple test app just loads a model, and also enables drag & drop to easily test other models. Inspector is always open, and fast refresh is supported so as you make changes to React components, you immediately see them in the test app, without reloading the page or re-creating the scene. You can run and debug the test app from VSCode by just running the new "Inspector v2 development (Chrome)" launch command. This enables a very tight dev loop. Hopefully in the future we can see what it would take to bring fast refresh to other tools, like Sandbox and Playground, but for now this is a low cost solution to help speed up Inspector v2 development.
- Remove redundant calls of normalize. - Fix type of normal.light. - Added `_throwIfDisposedOrAborted` to ensure timely termination if the parent operation is aborted.
|
WebGL2 visualization test reporter: |
|
Visualization tests for WebGPU |
packages/tools/viewer/src/viewer.ts
Outdated
| this._iblDirectionCalculationAbortController?.abort(new AbortError("New IBL dominant direction calculation initiated before previous calculation finished.")); | ||
| const iblDirectionCalculationAbortController = (this._iblDirectionCalculationAbortController = new AbortController()); | ||
|
|
||
| return await this._iblDirectionCalculationLock.lockAsync(async () => { |
There was a problem hiding this comment.
Oh dang, sorry, I didn't realize in my first review pass that this function is only called through the _updateShadows code path, which already has a lock and an abort controller. If that is really the only place it will be called, and this function is really just part of the implementation of _updateShadowMap (just factored out into its own function for readability), then it doesn't need its own abort controller and lock. It only needs those if it is potentially called from multiple call sites that are not under the same lock. Sorry for the confusion. :/
| return this._cachedIblDirection; | ||
| } | ||
|
|
||
| this._cachedIblDirection = null; |
There was a problem hiding this comment.
iblCdfGenerator already caches the dominant direction so calling findDominantDirection multiple times should just return right away without recalculating the direction.
There was a problem hiding this comment.
Another thing to keep in mind is that the direction returned here is unrotated. i.e. if the IBL has a rotationY value other than 0, this vector should be manually rotated to match the IBL rotation.
There was a problem hiding this comment.
Oh, and the magnitude of the direction returned is an indication of how dominant the direction is. So you could, in theory, use the magnitude to adjust the darkness and sharpness of the shadow.
A magnitude near 1.0 would indicate that the light is very strongly coming from one direction. A magnitude closer to 0.0 indicates that the light is coming from lots of directions.
There was a problem hiding this comment.
iblCdfGeneratoralready caches the dominant direction so callingfindDominantDirectionmultiple times should just return right away without recalculating the direction.
Yes i knew it, my main motivation for adding this._cachedDominantDirection into the viewer was to allow the _rotateShadowLightWithEnvironment function to remain synchronous (because I feel like it gets called a lot of times when it's plugged into a slider).
There was a problem hiding this comment.
Oh, and the magnitude of the direction returned is an indication of how dominant the direction is. So you could, in theory, use the magnitude to adjust the darkness and sharpness of the shadow. A magnitude near 1.0 would indicate that the light is very strongly coming from one direction. A magnitude closer to 0.0 indicates that the light is coming from lots of directions.
Oh nice thx for the info !
|
Visualization tests for WebGPU |
|
WebGL2 visualization test reporter: |
The iblLightStrength value is derived from the magnitude of the IBL's dominant light direction to enhance shadow realism by tweaking ShadowGenerator.darkness and ShadowGenerator.blurKernel.
|
WebGL2 visualization test reporter: |
|
Visualization tests for WebGPU |
…ylonJS#16737) I am working on a project that uses the ArcRotateCamera in Orthographic mode. I want to use the `ArcRotateCamera.zoomOn` function to zoom to a set of meshes. Unfortunately, the zoomOn function updates only the radius, not the orthographic extents. This PR updates the orthographic extents along with the radius when zooming onto mesh(es) in orthographic mode. Here is a demo playground: https://playground.babylonjs.com/#545IZ4 **Before:** https://github.com/user-attachments/assets/77b8516e-4ec8-40e0-a4d5-4f9ae0d4104d **After:** https://github.com/user-attachments/assets/876c3b68-3638-48a2-9b71-c909a4ae55ef
There's one color attribute declaration that is missing and will cause an error. --------- Co-authored-by: Kevin Blackburn-Matzen <matzen@adobe.com>
- [x] On end trigger - [x] On start trigger - [x] support worldMatrix from emitter - [x] Preview popup - [x] Flow map - [x] Expose emitter - [x] Sprite sheet - [x] Delay start - [x] Particle trigger - [x] Condition - [x] Simple condition - [x] Auto expendable blocks - [x] Basic update - [x] Snippet & loading API - [x] Teleport - [x] Elbow - [x] Custom Emitter Type - [x] Trigonometrics - [x] Converter  --------- Co-authored-by: David Catuhe <david@catuhe.com>
- Remove redundant calls of normalize. - Fix type of normal.light. - Added `_throwIfDisposedOrAborted` to ensure timely termination if the parent operation is aborted.
The iblLightStrength value is derived from the magnitude of the IBL's dominant light direction to enhance shadow realism by tweaking ShadowGenerator.darkness and ShadowGenerator.blurKernel.
…into auto-lighting
|
You have made possible changes to the playground. https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/PLAYGROUND/refs/pull/16735/merge/ The snapshot playground with the CDN snapshot (only when available): Note that neither Babylon scenes nor textures are uploaded to the snapshot directory, so some playgrounds won't work correctly. |
|
You have changed file(s) that made possible changes to the sandbox. https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/SANDBOX/refs/pull/16735/merge/ |
|
Visualization tests for WebGPU |
|
WebGL2 visualization test reporter: |
|
Should we close that one @ryantrem ? |
9 participants
By using the new function
findDominantDirection()of IBL shadows feature, this PR try to improves the orientation of the shadow map's light source.