SuperFlex: Deformable Superquadrics for Point Cloud Decomposition, 10000x Faster Than Marching Primitives
orlitany · x · 2026-09-12
A team from ETH Zurich, Stanford, IMPA and USI Lugano announced SuperFlex, an ECCV 2026 paper (poster #15) on superquadric decomposition of point clouds for 3D scene understanding.
- Method: Adds bending and tapering parameters to superquadrics for high-fidelity representation of complex geometries; a novel loss and optimization scheme refines initial decompositions; the model trains via self-supervised joint volumetric and surface losses, with optional per-object refinement.
- Key advance: Enables direct prediction from noisy, partial real-world point clouds — previously infeasible for methods requiring full point clouds.
- Results: On ShapeNet it significantly outperforms all learning- and optimization-based baselines (SQ, CSA, EMS, etc.) on IoU; the only comparable method, Marching Primitives, uses 4x more primitives and is 10000x slower.
More from Research
- Open-weights SUPlime beats pyannote's commercial diarization with 15.86 DER — solyarisoftware · 2026-09-12
- AI out-persuades world champion debaters, raises donations nearly 3x better than pros — ben_j_todd · 2026-09-12
- Shifting local token interactions from inference to training-time lookups seen as a scaling win — AccBalanced · 2026-09-12
- Open-source RL for large MoEs with zero train-infer mismatch, teaching Qwen3.6-35B-A3B to play Wordle — kastnerkyle · 2026-09-12
- ICLR 2026 paper MoM: multiple memory states fix linear models' recall weakness — kastnerkyle · 2026-09-12
- Harvard researcher turned a fruit fly brain connectome into a Bitcoin trading bot — Scobleizer · 2026-09-12