Skip to content

Source paper

DVD: Deterministic Video Depth Estimation with Generative Priors

Hongfei Zhang, Harold Haodong Chen, Chenfei Liao and 12 more

Hongfei Zhang, Harold Haodong Chen, Chenfei Liao, Jing He, Zixin Zhang, Haodong Li, Yihao Liang, Kanghao Chen, Bin Ren, Xu Zheng, Shuai Yang, Kun Zhou, Yinchuan Li, Nicu Sebe, Ying-Cong Chen

arXiv 2603.12250 · 202616 upvotes on Hugging Face

Research ideas from this paper (3)

Generated by AI. The paper’s authors did not write these.

AI summary

The paper introduces DVD, a novel framework that deterministically adapts pre-trained video diffusion models into single-pass depth regressors, overcoming both the stochastic hallucinations of generative models and the semantic ambiguity of discriminative models. By bypassing iterative sampling, DVD achieves state-of-the-art zero-shot video depth estimation with high temporal coherence, utilizing significantly less task-specific training data than leading baselines.

How it works (AI summary)

DVD operates by mapping an RGB sequence to depth deterministically in a single forward pass using a pre-trained video Diffusion Transformer (DiT) operating in a compressed VAE latent space. It repurposes a specific diffusion timestep as a 'structural anchor', utilizes Latent Manifold Rectification (LMR) to preserve sharp spatial and temporal differentials, and applies a least-squares global affine solver to seamlessly align overlapping windows during long-video inference.