OVGGT: O(1) Constant-Cost Streaming Visual Geometry Transformer

TL;DR

OVGGT introduces a resource-efficient, training-free framework for 3D reconstruction from streaming video, maintaining fixed memory and compute costs regardless of sequence length while achieving high accuracy.

Contribution

It combines Self-Selective Caching and Dynamic Anchor Protection to enable long-horizon streaming inference with constant resource usage, surpassing prior methods.

Findings

Processes arbitrarily long videos within fixed VRAM.

Achieves state-of-the-art 3D geometric accuracy.

Supports indoor, outdoor, and ultra-long sequences.

Abstract

Reconstructing 3D geometry from streaming video requires continuous inference under bounded resources. Recent geometric foundation models achieve impressive reconstruction quality through all-to-all attention, yet their quadratic cost confines them to short, offline sequences. Causal-attention variants such as StreamVGGT enable single-pass streaming but accumulate an ever-growing KV cache, exhausting GPU memory within hundreds of frames and precluding the long-horizon deployment that motivates streaming inference in the first place. We present OVGGT, a training-free framework that bounds both memory and compute to a fixed budget regardless of sequence length. Our approach combines Self-Selective Caching, which leverages FFN residual magnitudes to compress the KV cache while remaining fully compatible with FlashAttention, with Dynamic Anchor Protection, which shields coordinate-critical tokens from eviction to suppress geometric drift over extended trajectories. Extensive experiments on indoor, outdoor, and ultra-long sequence benchmarks demonstrate that OVGGT processes arbitrarily long videos within a constant VRAM envelope while achieving state-of-the-art 3D geometric accuracy. Project page: https://vaisr.github.io/OVGGT/ Code: https://github.com/VAISR/OVGGT