Repository navigation
Feat/to flame3d - #110
Merged
Merged
Feat/to flame3d#110
Conversation
hi-liang
commented
Oct 7, 2026
hi-liang
commented
Member
- Adds export for flame3d pipeline, with some python scripts to workshop the output format
- Sets all output meshes to include captured vertex colors. If a scan is not colorized, export will colorize first
Every export that includes the mesh (Nerfstudio, OBJ, PLY, USDZ) now carries per-vertex colors sampled from the captured frames, and never the save-time normals colors, which are an in-app preview only. - Exporting a mesh format colorizes any uncolored scan first (ScanExportManager.colorizeForExport -> ScanPostprocessor.run with colorize, the Color button's pass). All six export/upload call sites await it, then read isColored on main for prepareExport. - prepareExport(vertexColorsFromCapture:) gates colors on isColored and defaults to false, so a caller that forgets it ships geometry only. Before this, PLY baked colors.bin unconditionally, shipping the normals preview for scans never colorized. - OBJ and the Nerfstudio mesh.obj write `v x y z r g b` (MeshConverter.objWithVertexColors); USDZ converts from that colored OBJ; PLY takes colors.bin. Colors ship only with exactly one per vertex. - Colorize already skips person-mask regions (and maskless frames) with Privacy Filter on, so no person is baked into the colors. - Docs: README, REQUIREMENTS, schemas, PRIVACY, APPSTORE, User Guide.
New export format that writes flame3d-core's `polycam` input directly, so a scan uploads to flame3d with no desktop conversion. Flame3DExport mirrors tools/scan4d-to-flame3d (the contract and reference implementation): - keyframes/images: stream frames + hi-res stills at one resolution (stills resized to the stream size, intrinsics scaled); cube faces, odd-aspect stills and stems without a unique frame number left out, since flame3d's SAM3 runs need one resolution and frame-numbered stems. - keyframes/corrected_cameras: Polycam camera records (raw-frame c2w). - keyframes/depth: native LiDAR depth, zero maps where a frame has none. - raw.glb: mesh.obj + captured vertex colors (COLOR_0, linear); the export colorizes first like every mesh export. - mesh_info.json: registration as alignmentTransform. - Built from the staged Polycam payload, so the privacy passes have run. - Own store-only zip with entries at the root: the directory zip puts the folder on top, and flame3d extracts uploads as-is. tools: `scan4d_to_flame3d.py --verify` checks any flame3d bundle (layout, frame sets, sizes, depth, GLB, alignment vs LiDAR) before upload, for checking the app's export off-device.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.