arXiv Computer Vision

Latent Riemannian Flow Matching for Geometry-Grounded 3D Foundation Models

The paper introduces Latent Riemannian Flow Matching, a method that performs flow matching directly in the latent space of geometric foundation models like VGGT. By respecting the product of zero‑mean hyperspheres imposed by LayerNorm, the approach avoids the pitfalls of treating the latent space as unconstrained Euclidean. Experiments on RealEstate10K, ScanNet++, and ETH3D show that this technique outperforms recent scene‑generation baselines in both novel‑view appearance and aggregated 3D geometry, and it transfers to other pretrained 3D backbones such as Depth Anything 3.

arXiv Computer Vision
Oct 16

ControlFlow3D: Distilling Multi-View Knowledge into Latent Flow Matching for Point Cloud Upsampling

ControlFlow3D is a framework that upsamples point clouds by distilling knowledge from pretrained foundation models into a lightweight generation network. It performs conditional flow matching in a latent space defined by a frozen 3D encoder, using unsupervised soft KMeans to cluster multi‑view image tokens and 3D latent tokens into shared prototypes that guide the flow trajectory. A Mamba‑based geometry decoder then maps latent features to dense coordinates with linear complexity, and the authors introduce ShapeNetPU, a 35K‑object multi‑modal benchmark 31 times larger than previous datasets.

By Yuang Liu, Zhi Zuo, Zhengkai Zhao, Lirui Zhang, Pan Gao, Hao Feng, Zhengzhe Liu
arXiv Computer Vision
Sep 22

GAE: Learning a Geometry-Native Latent Space for 3D-Consistent World Generation

The paper introduces the Geometry‑Native Autoencoder (GAE), a compact latent space that can be decoded into appearance, depth, camera parameters, and point maps, enabling 3D‑consistent world generation. By reparameterizing a geometry foundation model’s features, GAE replaces traditional appearance‑centric latents and improves visual quality and 3D coherence, achieving significant reductions in FVD and camera‑trajectory error on benchmark datasets. The work demonstrates that a geometry‑native latent space can serve as a shared interface between perception and generation models.

By Jiahao Lu, Minghao Yin, Wenbo Hu, Hengyu Liu, Wang Zhao, Sai-Kit Yeung, Ying Shan, Yuan Liu
arXiv AI
Oct 16

Motion- and Gravity-Aware Feedforward 3D Reconstruction

The paper presents a motion- and gravity-aware framework for feed-forward transformer models in multi-view 3D reconstruction. By providing residual optical flow and gravity direction as privileged signals during training, the model learns stronger geometric representations while remaining efficient at inference. Experiments demonstrate consistent accuracy gains, particularly in dynamic scenes and challenging camera motions, with the optimal integration stage depending on the backbone architecture.

By Shiyao Qian, Yuan Ren, Dongfeng Bai
arXiv Computer Vision
Sep 24

GaussianDS: Depth-supervised Semantic Gaussian Splatting for Scene Understanding

GaussianDS introduces a depth‑supervised framework for 3D Gaussian Splatting that jointly optimizes RGB appearance, depth, and compact semantics from scratch. By arranging multi‑view images into a pose‑aware pseudo‑video and propagating view‑consistent masks via SAM2, the method aligns semantic lifting with geometric cues, using depth supervision and edge‑aware refinement to curb semantic drift and boundary leakage. The approach achieves state‑of‑the‑art performance on LERF and 3D‑OVS benchmarks while preserving high‑fidelity reconstruction and enabling downstream tasks such as 3D object removal.

By Yufei Zhang, Chenlu Zhan, Hongwei Wang
arXiv AI
Oct 19

Scalable Adaptation of 3D Geometric Foundation Models via Weak Supervision from Internet Video

The paper introduces SAGE, a framework that adapts 3D geometric foundation models using weak supervision from raw Internet videos. SAGE employs a hierarchical mining pipeline to generate informative training trajectories, sparse geometric anchoring via SfM point clouds, and dense differentiable consistency through 3D Gaussian rendering, while also replaying a small amount of supervised data to avoid catastrophic forgetting. Experiments on two different backbones show that SAGE reduces Chamfer Distance on unseen benchmarks by 20–42% and consistently improves performance, demonstrating that Internet video can serve as a scalable source for geometric learning.

By Zihui Gao, Ke Liu, Donny Y. Chen, Duochao Shi, Guosheng Lin, Jia-Wang Bian, Hao Chen, Chunhua Shen
arXiv Computer Vision
Sep 7

Learning 3D Editing without Paired Supervision via Generative Prior Distillation

The paper introduces a framework for instruction‑guided 3D editing that does not require paired 3D supervision. It distills visual, semantic, and geometric knowledge from foundation models into a 3D editing model using a differentiable rendering pipeline, guided by a 2D visual prior from an image editing model and a semantic prior from a Vision‑Language Model. A 3D‑aware Distribution Matching regularization is added to prevent geometric collapse and ensure realistic 3D outputs, leading to superior instruction fidelity and cross‑view consistency compared to state‑of‑the‑art baselines.

By Hao Wen, Weibin Yun, Hongxing Fan, Haotian Lu, Rui Chen, Zehuan Huang, Lu Sheng
arXiv Computer Vision
Oct 21

Latent Bridge: From Video Latents to 3D Geometry via Feature Alignment

Latent Bridge connects video generative latent spaces with geometric feature spaces, enabling direct prediction of 3D geometry—cameras, depth, and point clouds—without an intermediate RGB decoding step. The method trains representation‑specific bridges for Wan2.1 and V‑RAE via feature distillation on RGB videos while keeping the video encoder and geometry model frozen, requiring no ground‑truth geometry. At inference, the same video latents support both RGB decoding and geometry prediction, and the trained bridge can be reused across image‑to‑video and text‑to‑video generation, achieving competitive reconstruction quality on Co3Dv2 and WildRGB‑D with efficient end‑to‑end generation.

By Shuai Zhang, Zhanpeng Ouyang, Zhengqing Chen, Ruohong Mei, Shaoyu Chen, Weiqiang Ren, Hui Sun, Qian Zhang, Wenyu Liu, Xinggang Wang
arXiv Computer Vision
Sep 16

Bi-FlowGS: Bridging Generative View Completion and Gaussian Geometry through Bidirectional Flow Co-Refinement

Bi-FlowGS introduces a bidirectional co-refinement framework that links generative view completion with 3D Gaussian Splatting geometry. It employs Video-to-Geometry Flow Distillation (V2G) to transfer temporal correspondence from restored videos into Gaussian geometry, mitigating the Geometry Cheating problem. Simultaneously, Geometry-to-Video Flow-Guided Restoration (G2V) uses the current 3DGS geometry to guide temporally consistent video restoration, creating a loop where restored videos and optimized geometry iteratively improve each other, leading to better rendering quality and geometric consistency on wide-baseline and 360° benchmarks.

By Yuetong Wang, Jinsheng Quan, Yi Yang, Yawei Luo