Online Dense Point Tracking with Streaming Memory

2025-03-09

AI Summary

Online dense point tracking requires persistent localization of all points from the first frame throughout a video sequence, particularly under occlusions and appearance variations. Conventional optical flow methods neglect temporal consistency, leading to drift, while existing sliding-window approaches suffer from inefficient propagation and poor long-range tracking performance. This paper introduces SPOT, a lightweight end-to-end differentiable framework for online inference. SPOT pioneers a streaming memory mechanism that jointly integrates feature-enhanced memory retrieval, sensory short-term motion modeling, and visibility-guided splatting-based propagation. On the CVO benchmark, SPOT achieves state-of-the-art accuracy; on TAP-Vid and RoboTAP, it matches or surpasses offline models despite operating online. SPOT reduces model parameters by 10× and accelerates inference by over 2× compared to prior online methods.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Time-Series/Data Streams
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Dense point tracking is a challenging task requiring the continuous tracking of every point in the initial frame throughout a substantial portion of a video, even in the presence of occlusions. Traditional methods use optical flow models to directly estimate long-range motion, but they often suffer from appearance drifting without considering temporal consistency. Recent point tracking algorithms usually depend on sliding windows for indirect information propagation from the first frame to the current one, which is slow and less effective for long-range tracking. To account for temporal consistency and enable efficient information propagation, we present a lightweight and fast model with extbf{S}treaming memory for dense extbf{PO}int extbf{T}racking and online video processing. The extbf{SPOT} framework features three core components: a customized memory reading module for feature enhancement, a sensory memory for short-term motion dynamics modeling, and a visibility-guided splatting module for accurate information propagation. This combination enables SPOT to perform dense point tracking with state-of-the-art accuracy on the CVO benchmark, as well as comparable or superior performance to offline models on sparse tracking benchmarks such as TAP-Vid and RoboTAP. Notably, SPOT with 10$ imes$ smaller parameter numbers operates at least 2$ imes$ faster than previous state-of-the-art models while maintaining the best performance on CVO. We will release the models and codes at: https://github.com/DQiaole/SPOT.

Problem

• Addresses challenges in dense point tracking in videos.
• Improves temporal consistency and information propagation efficiency.
• Enhances tracking accuracy and speed with fewer parameters.

Innovation

• Streaming memory for efficient point tracking
• Customized memory reading enhances feature accuracy
• Visibility-guided splatting improves information propagation