FutureDepth: Learning to Predict the Future Improves Video Depth Estimation

2024-03-19

European Conference on Computer Vision

AI Summary

To address insufficient motion modeling and suboptimal exploitation of multi-frame information in video depth estimation, this paper proposes FutureDepth—a prediction-driven spatiotemporal representation learning framework. Its core innovation lies in jointly incorporating a future-frame feature prediction network (F-Net) and an adaptive masked multi-frame feature reconstruction network (R-Net), which implicitly model inter-frame motion and correspondence through iterative forward prediction and masked self-encoding. The method integrates temporal prediction, multi-frame reconstruction, depth decoding, and refinement modules into a unified architecture. Evaluated on diverse benchmarks—including NYUDv2, KITTI, DDAD, and Sintel—FutureDepth achieves state-of-the-art accuracy, significantly outperforming existing video-based depth estimation methods. Notably, its inference efficiency matches that of monocular models, enabling practical deployment without sacrificing performance.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

In this paper, we propose a novel video depth estimation approach, FutureDepth, which enables the model to implicitly leverage multi-frame and motion cues to improve depth estimation by making it learn to predict the future at training. More specifically, we propose a future prediction network, F-Net, which takes the features of multiple consecutive frames and is trained to predict multi-frame features one time step ahead iteratively. In this way, F-Net learns the underlying motion and correspondence information, and we incorporate its features into the depth decoding process. Additionally, to enrich the learning of multiframe correspondence cues, we further leverage a reconstruction network, R-Net, which is trained via adaptively masked auto-encoding of multiframe feature volumes. At inference time, both F-Net and R-Net are used to produce queries to work with the depth decoder, as well as a final refinement network. Through extensive experiments on several benchmarks, i.e., NYUDv2, KITTI, DDAD, and Sintel, which cover indoor, driving, and open-domain scenarios, we show that FutureDepth significantly improves upon baseline models, outperforms existing video depth estimation methods, and sets new state-of-the-art (SOTA) accuracy. Furthermore, FutureDepth is more efficient than existing SOTA video depth estimation models and has similar latencies when comparing to monocular models

Problem

• Depth Estimation
• Video Analysis
• Multi-frame Information

Innovation

• FutureDepth
• F-Net
• R-Net