SpatialBench evaluates 41 spatial foundation models across 6 paradigms and 5 task suites, finds they are not all-round players, and introduces the DA-Next-5M dataset plus DA-Next baseline model.
hub Canonical reference
FastVGGT: Training-Free Acceleration of Visual Geometry Transformer
Canonical reference. 86% of citing Pith papers cite this work as background.
35
Pith papers citing it
Background
86% of classified citations
abstract
Foundation models for 3D vision have recently demonstrated remarkable capabilities in 3D perception. However, scaling these models to long-sequence image inputs remains a significant challenge due to inference-time inefficiency. In this work, we present a detailed analysis of VGGT, a state-of-the-art feed-forward visual geometry model and identify its primary bottleneck. Visualization further reveals a token collapse phenomenon in the attention maps. Motivated by these findings, we explore the potential of token merging in the feed-forward visual geometry model. Owing to the unique architectural and task-specific properties of 3D models, directly applying existing merging techniques proves challenging. To this end, we propose FastVGGT, which, for the first time, leverages token merging in the 3D domain through a training-free mechanism for accelerating VGGT. we devise a unique token partitioning strategy tailored to 3D architectures and tasks, effectively eliminating redundant computation while preserving VGGT's powerful reconstruction capacity. Extensive experiments on multiple 3D geometry benchmarks validate the effectiveness of our approach. Notably, with 1000 input images, FastVGGT achieves a 4x speedup over VGGT while mitigating error accumulation in long-sequence scenarios. These findings underscore the potential of token merging as a principled solution for scalable 3D vision systems. Code is available at: https://mystorm16.github.io/fastvggt/.
hub tools
citation-role summary
background 6
baseline 1
citation-polarity summary
years
2026 35representative citing papers
DrivingDepth achieves SOTA metric depth on nuScenes by residual pixel-wise scale correction on frozen foundation models using sparse LiDAR prompts, preserving geometric consistency.
CasaMaestro predicts metric depth and poses from sparse multi-view panoramas to enable fast house-scale 3D reconstruction.
PaceVGGT reduces VGGT inference latency by up to 5.1x on ScanNet-50 via pre-AA token pruning with a distilled Token Scorer, per-frame keep budgets, adaptive merge/prune, and feature-guided restoration, while preserving reconstruction quality on ScanNet-50 and 7-Scenes.
Ground4D resolves temporal conflicts in feedforward 4D Gaussian reconstruction for off-road scenes via voxel-grounded temporal aggregation with intra-voxel softmax and surface normal regularization, outperforming prior methods on ORAD-3D and RELLIS-3D while generalizing zero-shot.
RobotPan predicts metric-scaled compact 3D Gaussians from calibrated multi-view inputs via spherical coordinates and hierarchical voxel priors for real-time 360° robotic perception and reconstruction.
Mem3R achieves better long-sequence 3D reconstruction by decoupling tracking and mapping with a hybrid memory of TTT-updated MLP and explicit tokens, reducing model size and trajectory errors.
VGGT-360 delivers geometry-consistent zero-shot panoramic depth by converting panoramas into multi-view 3D reconstructions via VGGT models and three plug-and-play correction modules, then reprojecting the result.
STAC compresses KV caches in streaming 3D reconstruction transformers via temporal token preservation with decayed attention, spatial voxel compression, and chunked multi-frame optimization, delivering 10x memory reduction and 4x faster inference at SOTA quality.
ZipMap achieves linear-time bidirectional 3D reconstruction by zipping image collections into a compact stateful representation via test-time training layers.
FFAvatar uses a Transformer-based 3D Gaussian model with alternating attention and sparse-to-dense learning to enable feed-forward, incremental reconstruction of animatable 4D head avatars from sparse portrait images.
EPS3D is an end-to-end architecture for 3D panoptic segmentation from multi-view images that uses distillation and semantic-instance mutual enhancement to achieve higher benchmark performance and speed than prior methods.
Anchor3R reframes feed-forward 3D reconstruction as current-centric local measurement prediction, using loop-closure and motion averaging to produce coherent global maps from visual streams.
KeyVT improves zero-shot 3D question answering by hierarchically selecting semantically and geometrically relevant views and using optimal transport to extract representative tokens from them.
DeblurNVS restores geometric representations via latent diffusion to enable high-fidelity novel view synthesis directly from sparse motion-blurred inputs.
RayDer is a unified transformer backbone for self-supervised static-scene novel view synthesis that absorbs dynamic content as a nuisance factor and shows power-law scaling with data and compute while matching supervised methods in zero-shot settings.
UniT unifies online and offline 3D geometry perception via a Group Autoregressive Transformer that processes observation groups with anchor-free point map prediction and a scale-adaptive loss.
A closed-form scalar frame-level gate α_t derived from internal feature changes extends effective memory in recurrent 3D reconstruction and improves accuracy on long sequences up to 4541 frames.
FGQ applies diagonal Fisher information to guide learnable affine transformations in PTQ for multi-task VGGT, yielding up to 39% relative gains over baselines at 4-bit quantization.
A training-free progressive decoupling framework improves dynamic depth estimation in 4D reconstruction via mask-guided pose decoupling, topological subspace surgery, and Bayesian fusion, yielding better point-cloud metrics on benchmarks.
RetrieveVGGT enables constant-memory long-context streaming 3D reconstruction by retrieving relevant frames via query-key similarities in VGGT's first attention layer, outperforming StreamVGGT and others.
Spark3R achieves up to 28x speedup on 1000-frame 3D reconstruction inputs by asymmetrically reducing query and key-value tokens in Vision Transformers while keeping competitive quality.
SpatialFusion internalizes 3D geometric awareness into unified image generation models by pairing an MLLM with a spatial transformer that produces depth maps to constrain diffusion generation.
ELSA casts online softmax attention as a prefix scan over monoid (m,S,W) to deliver exact FP32 semantics, O(n) memory, O(log n) depth, and Tensor-Core independence as a drop-in kernel.
citing papers explorer
-
Mem3R: Streaming 3D Reconstruction with Hybrid Memory via Test-Time Training
Mem3R achieves better long-sequence 3D reconstruction by decoupling tracking and mapping with a hybrid memory of TTT-updated MLP and explicit tokens, reducing model size and trajectory errors.
-
4DVGGT-D: 4D Visual Geometry Transformer with Improved Dynamic Depth Estimation
A training-free progressive decoupling framework improves dynamic depth estimation in 4D reconstruction via mask-guided pose decoupling, topological subspace surgery, and Bayesian fusion, yielding better point-cloud metrics on benchmarks.
-
Attention Itself Could Retrieve.RetrieveVGGT: Training-Free Long Context Streaming 3D Reconstruction via Query-Key Similarity Retrieval
RetrieveVGGT enables constant-memory long-context streaming 3D reconstruction by retrieving relevant frames via query-key similarities in VGGT's first attention layer, outperforming StreamVGGT and others.
-
SpatialFusion: Endowing Unified Image Generation with Intrinsic 3D Geometric Awareness
SpatialFusion internalizes 3D geometric awareness into unified image generation models by pairing an MLLM with a spatial transformer that produces depth maps to constrain diffusion generation.
-
Feed-Forward 3D Scene Modeling: A Problem-Driven Perspective
The paper proposes a problem-driven taxonomy for feed-forward 3D scene modeling that groups methods by five core challenges: feature enhancement, geometry awareness, model efficiency, augmentation strategies, and temporal-aware modeling.
-
Scal3R: Scalable Test-Time Training for Large-Scale 3D Reconstruction
Scal3R achieves better accuracy and consistency in large-scale 3D scene reconstruction by maintaining a compressed global context through test-time adaptation of lightweight neural networks on long video sequences.