model: image to 3drendering - #533
Open
rebel-ChanyongJeon wants to merge 7 commits into
Open
Conversation
…ruction - Add RBLNTripoSRForImageTo3D and RBLNTripoSRForImageTo3DConfig - Implement triplane-based scene code generation and mesh extraction - Support NPU triplane query with CPU fallback - Add configuration, architecture, and modeling modules - Register TripoSR in optimum-rbln transformers model registry
- Add RBLNSF3DForImageTo3D (Stable Fast 3D): architecture port, NPU compile wrappers, marching-cubes mesh extraction with vertex colors, lazy exports - Register SF3D in optimum.rbln / transformers / models __init__ - TripoSR: top-level imports (einops, trimesh, skimage), marching_cubes in architecture; align get_pytorch_model, _update_rbln_config, _create_runtimes with RBLNModel; drop @torch.no_grad on forward - pyproject: add einops, trimesh, scikit-image, omegaconf as core dependencies - Refresh sf3d_README / triposr_README (pipeline, deps table)
rebel-ChanyongJeon
changed the title
rebel-ChanyongJeon
force-pushed
the
feat/image-to-3drendering
branch
from
March 20, 2026 07:33
8bc8e6c to
735df6a
Compare
- Sort SF3D/TripoSR in lazy __init__ and TYPE_CHECKING imports - modeling_sf3d: remove unused total_out, zip(..., strict=True) - sf3d_architecture: consolidate imports, B007 use _i - triposr_architecture: C416 dict(enumerate), keep chunk_length - cli: noqa PLC0415 for lazy optimum.rbln import
- TestImageTo3DModelClasses: get_rbln_config_class smoke tests (always on) - TestTripoSRForImageTo3D / TestSF3DForImageTo3D: BaseTest.TestModel, FULL level - RBLN_AUTO_CLASS None until Auto mapping exists
…deps - After editable install, uv pip install trimesh, scikit-image, einops, omegaconf so test_transformers can import SF3D/TripoSR without ModuleNotFoundError - Add same packages to dependency-groups.tests for uv lock / sync alignment
SciPy 1.17+ wheels (manylinux_2_27/2_28) caused uv sync failures in CI (extract/I-O timeouts). Keep scipy 1.15.x via constraint-dependencies and lockfile so frozen installs match.
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.
Pull Request Description
Type of Change
Changes Overview
RBLNTripoSRForImageTo3D): triplane-based single-image 3D mesh reconstruction, NPU triplane query with CPU-side mesh extraction path, config / architecture / modeling modules andtriposr_README.RBLNSF3DForImageTo3D): architecture port, NPU compile/runtime integration, marching-cubes mesh extraction with vertex colors, lazy exports,sf3d_README.optimum.rblntransformers model registry (__init__.pyexports).pyproject.toml):einops,trimesh,scikit-image,omegaconffor image-to-3D pipelines and mesh post-processing.get_pytorch_model/_update_rbln_config/_create_runtimeswith RBLN conventions; top-level imports for heavy deps; remove@torch.no_gradonforward; refreshtriposr_README.get_pytorch_model/_update_rbln_config/_create_runtimes; NPU triplane decoder + CPU fallback; marching cubes inextract_meshwith vertex colors; top-leveleinops/trimesh/scikit-image; no@torch.no_gradonforward(mesh extraction usesextract_mesh);sf3d_READMEwith pipeline and dependency table.Motivation and Context
Expose image-to-3D reconstruction on RBLN NPU for TripoSR and SF3D so users can compile and run these pipelines via
optimum-rbln, consistent with otherRBLN*For*model integrations.Related Issues