UniForge3D From PKU and Others Hits SOTA in 3D Asset Generation From Sparse Pose-Free Images
PekingUniversity · hf · 2026-10-14
A team from Peking University, Math Magic, and HKUST releases UniForge3D, tackling complete 3D asset generation from sparse, camera-pose-free images—preserving observed geometry while completing unseen regions:
- Shared anchor-guided space: unifies reconstruction and generation in a single space, resolving the mismatch between camera-relative and canonical generation coordinate systems; the space follows the anchor view's azimuth and keeps the object upright
- Bidirectional feature exchange: reconstruction and generation branches exchange features, predicting per-view point maps and a complete sparse voxel structure respectively
- View-selective conditioning: point map prediction is conditioned on the generated voxel structure to establish pixel-voxel correspondences, and the most informative input view is selected for each voxel to guide detail generation while preserving observed geometric details
It achieves state-of-the-art appearance and geometry results on the Toys4K and GSO benchmarks.
More from Research
- Chollet: Ideal AI-for-Science Outcome Is Millions of AI-Aided Scientists, Not a Few Rich Labs — fchollet · 2026-10-14
- Feeling a dipole's electric field along an ellipse: 2.08kq at the end, 0.13kq on top — TinfoilTricorn · 2026-10-14
- Alibaba's Representation Guidance: plain pixel-diffusion Transformers beat bigger models — bdsqlsz · 2026-10-14
- Visualizing the Fourier series of a sawtooth wave, with Gibbs overshoot hitting ~9% — TinfoilTricorn · 2026-10-14
- Gemma Embedding 2 hits SOTA on Vietnamese speech after 3,000 hours of audio, 2 hours on 8xH100 — stablequan · 2026-10-14
- KILVO: robust multimodal odometry for humanoid robots fusing encoders, IMU, LiDAR and vision — rsasaki0109 · 2026-10-14