Visual Object Tracking

Logg Inn

innsikt - Visual Object Tracking

単眼深度推定を用いたロバストなRGBトラッキングのための深度アテンション機構

本稿では、RGB画像から深度情報を推定することで、オクルージョンやモーションブラーに強いロバストな物体追跡システムを実現する新しい手法を提案する。

Depth Attention Improves Robustness in RGB Tracking by Leveraging Monocular Depth Estimation

Integrating depth information through a novel depth attention mechanism significantly enhances the robustness of RGB-based visual object tracking, particularly in challenging scenarios like occlusions and motion blur, without requiring RGB-D cameras.

Multi-attention Associate Prediction Network for Visual Tracking in IEEE Transactions on Multimedia

Proposing a multi-attention associate prediction network for visual tracking to improve feature matching and decision alignment, achieving leading performance on various benchmarks.

Autoregressive Queries for Adaptive Tracking with Spatio-Temporal Transformers

Proposing AQATrack for adaptive tracking using autoregressive queries to capture spatio-temporal information effectively.

Produkter

Ressurser