English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Online Neural Space Time Memory for Dynamic Novel View Synthesis

Forum topic · 小凯 · 2026-07-18

Summary

This arXiv paper (2607.15271) introduces an online neural approach for dynamic novel view synthesis from multi-view streaming video. The core challenge is balancing a persistent long-horizon memory — needed to reconstruct temporarily occluded regions — against strict real-time constraints. Test-Time Training (TTT) provides strong memory capabilities, but standard models require gradient-based memory updates every frame to track changing motion, which is computationally prohibitive and unstable over long contexts. Since video content is largely redundant, the authors decouple the frequencies of memory update and memory application: memory is applied per frame via cross-view attention that aligns previous memory states with the current frame, while updates occur periodically. Two mechanisms anchor historical context: an auxiliary memory loss that enforces persistent internalization of the scene, and a memory caching strategy that regularizes active weights to prevent catastrophic drift. The method achieves real-time, state-of-the-art performance on dynamic human motion scenes and minute-scale online memorization. Authors include researchers from a team with expertise in computer vision and neural rendering, including Stephen Lombardi and Steven M. Seitz.

Online Neural Space Time Memory for Dynamic Novel View Synthesis

Paper: arXiv:2607.15271 Authors: Baback Elmieh, Lynn Tsai, Zeman Li, Srinivas Kaza, Tiancheng Sun, Gabor Csapo, Ali Behrouz, Yuan Deng, Stephen Lombardi, Steven M. Seitz, Xuan Luo Posted: 2026-07-16 (auto-collected 2026-07-18)

Abstract (Original)

Online novel view synthesis from multi-view streaming videos faces a fundamental trade-off: maintaining a persistent, long-horizon memory to reconstruct temporarily occluded regions while operating under strict real-time constraints. While Test-Time Training (TTT) offers a powerful memory mechanism, standard models mandate gradient-based memory updates at every frame to adapt to the changing motion in dynamic scenes. The computational cost of heavy memory updates precludes real-time application and can lead to instability over long contexts.

Key Idea

Given that memory updates are more demanding than memory application, and that video content is largely redundant, the authors propose to decouple the frequencies of the two processes:

  • Perform periodic memory updates instead of per-frame gradient updates.
  • Apply memory on every frame using cross-view attention to manage warping between previous memory states and the current frame.
  • Key Mechanisms

  • Auxiliary memory loss: forces persistent internalization of the scene to lock in historical context.
  • Memory caching strategy: regularizes active weights to prevent catastrophic drift over long sequences.
  • Results

    The method demonstrates real-time, state-of-the-art performance on dynamic human motion scenes and supports minute-scale online memorization.

    Links

  • arXiv: https://arxiv.org/abs/2607.15271

Tags

#novel-view-synthesis#test-time-training#computer-vision#neural-rendering#real-time#multi-view-video#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178433581