- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper presents PartSynth, a four-stage image-to-3D framework built upon TRELLIS.2 that jointly generates compositional multi-part geometry and part-aware PBR materials. Its core representation, Sparse Part Structure (SPS), decomposes a coarse whole-shape scaffold into up to set-indexed volumes via graph coloring on the part-contact graph, compressed onto a shared sparse latent support .
Critical strengths
-
This work makes a contribution to pushing native 3D compositional generation forward. By instantiating set-indexed slots only on the active sparse scaffold support ( with ), it provides a clean and scalable paradigm for expanding the capacity of PartCrafter/PartPacker-style generators without suffering from dense volumetric memory explosion. This formulation handles complex non-bipartite contact graphs and overlapping parts in a single fast forward pass (~10s). Although the current implementation has limitations in per-part resolution and controllability, the underlying SPS representation offers valuable insights for the 3D generation community and justifies acceptance.
-
Extending pretrained monolithic geometry and PBR flow models with 4D rotary positional encoding enables global information exchange across all parts while keeping set-level surface decoding separated. As verified by the
PartSynth-PerSetablation and the proposed PI-ME metric, this design effectively reduces cross-part material bleeding while maintaining whole-shape coherence.
Critical weaknesses
-
The geometric quality of individual parts in the supplementary material and normalized part galleries (Fig. 3) is noticeably coarse and low-resolution. The pipeline stops at the stage and omits the second-stage high-resolution refinement of TRELLIS.2 (), likely due to the sequence-length overhead of concatenating up to sets in joint self-attention. Because all parts must share a single global grid rather than being synthesized in per-part full-resolution grids (as in FullPart or OmniPart), small components receive very few active voxels. Moreover, part topology is permanently locked on the sub-scaffolds (after a pass through ), making the pipeline highly sensitive to quantization errors that cause thin parts to fragment or nearby parts in the same set to merge irreversibly.
-
Following an unguided end-to-end paradigm like PartCrafter rather than a 2D-controllable framework like OmniPart leaves the part decomposition process as an uncontrollable black box. Conditioned only on a single RGB image, the model provides no mechanism for users to specify part boundaries, guide where splits should occur, or even control the decomposition granularity (e.g., whether to split an object into a few coarse assemblies or dozens of fine sub-components).
Questions for authors
-
Can the 4D joint-attention architecture be extended to TRELLIS.2's second-stage cascade refinement () to improve per-part geometric sharpness, or does the -fold token expansion make structured latents computationally prohibitive?
-
Since SPS currently operates purely from a single image without structural guidance, how could user-specified 2D part masks (as in OmniPart) or granularity controls be injected into the slot generation process so that users can control how parts are partitioned?
Thanks for the detailed rebuttal. My concerns about inference time and training cost are resolved. One small request: please reconcile the rebuttal timings with the per-stage times in Section 4.1.
Remaining concerns:
I am still borderline.