feat(graphene): multiscale meshes #460
Conversation
| @@ -3,272 +3,67 @@ name: Build | |||
| on: [push, pull_request] | |||
|
|
|||
| jobs: | |||
| client: | |||
There was a problem hiding this comment.
I don't think you intended to include the changes in this file.
| @@ -0,0 +1,32 @@ | |||
| runtime: python27 | |||
There was a problem hiding this comment.
Probably not intended to be included in this PR.
| @@ -0,0 +1,6 @@ | |||
| { | |||
src/neuroglancer/mesh/draco/index.ts
Outdated
| @@ -68,7 +68,7 @@ export async function decodeDracoPartitioned( | |||
| heap.set(buffer, offset); | |||
| numPartitions = partition ? 8 : 1; | |||
| const code = (m.instance.exports.neuroglancer_draco_decode as Function)( | |||
| offset, buffer.byteLength, partition, vertexQuantizationBits, true); | |||
| offset, buffer.byteLength, partition, vertexQuantizationBits, false); | |||
There was a problem hiding this comment.
This will need to be an option, since we don't want to change this option for the neuroglancer precomputed datasource.
There was a problem hiding this comment.
Yes I will do that when the PR is ready to merge. This was just to get the meshes to appear during development.
| @@ -44,14 +44,15 @@ import { SharedWatchableValue } from 'neuroglancer/shared_watchable_value'; | |||
| import { DisplayDimensionRenderInfo } from 'neuroglancer/navigation_state'; | |||
| import { forEachVisibleSegment } from 'neuroglancer/segmentation_display_state/base'; | |||
| import { computeChunkBounds } from 'neuroglancer/sliceview/volume/backend'; | |||
| import { verifyObject } from 'src/neuroglancer/util/json'; | |||
There was a problem hiding this comment.
Should be 'neuroglancer/util/json' -- no 'src/' prefix.
| lodScales: new Float32Array(response.lodScales), | ||
| octree: new Uint32Array(response.octree), | ||
| vertexOffsets: new Float32Array(response.lodScales.length * 3), | ||
| clipLowerBound: vec3.create(), |
There was a problem hiding this comment.
This is why the meshes aren't displaying properly --- the mesh rendering code calculates bounds for each octree node by calculating the bounds of the node according to the grid, and then restricting them by these overall clip bounds. If the resultant bounding box does not intersect the view frustum, then the octree node won't be rendered. With these values for the clip bounds, all of the calculated bounds end up being (0, 0, 0)---(0,0,0), which both defeats the normal logic of selectively rendering only visible chunks, and also means if the view frustum does not contain (0, 0, 0) then nothing will be rendered.
There was a problem hiding this comment.
If you don't have the real bounds available, but you can just calculate the bounds that include all of the fragments present.
There was a problem hiding this comment.
Thanks! I have pushed changes and updated the chunkedgraph deployment to return clip bounds.
Realized that zoom-out disappearance is being caused by resolution, changing it fixes that. But this means higher resolution meshes are loaded earlier than expected.
Currently chunkedgraph is passing bounding box of the dataset for clip lower and upper bounds. Should that fix the above issue? It doesn't look like it does though, I'm not familiar with webGL.
akhileshh
force-pushed
the
graphene-multiscale-meshes
branch
from
23a25a6
to
be12f35
Compare
akhileshh
force-pushed
the
graphene-multiscale-meshes
branch
from
be12f35
to
88a1b62
Compare
No description provided.