GOT-10k: A Large High-diversity Benchmark and Evaluation Platform for Single Object Tracking
Visual Object Tracking & Evaluation Technology & Large High-diversity Benchmark
GOT-10k is constructed to evaluate the generalization ability of trackers on unseen object classes and motion patterns. The platform provides a high-quality video trajectory dataset containing 10,000 video segments, 563 object classes, 87 motion patterns, and 1.5 million tight annotations.
GOT-10k is the supporting platform for a research accepted by IEEE TPAMI 2019. As of Feb. 2024, the platform has received 3.24M+ page views, 6k+ downloads, 18k+ trackers from 160+ countries and regions worldwide.
VideoCube: A Large-scale Multi-dimensional Global Instance Tracking Intelligent Evaluation Platform
Visual Object Tracking & Large-scale Benchmark Construction & Intelligent Evaluation Technology
This work builds on the concept of human-like modeling and expands the definition of single object tracking (SOT) task. It presents a new task called global instance tracking (GIT). This work builds a large-scale, multi-dimensional global instance tracking task intelligent evaluation platform called VideoCube.
VideoCube is the supporting platform for research accepted by IEEE TPAMI 2022. As of Feb. 2024, the platform has received 315k+ page views, 1k+ downloads, 400+ trackers from 130+ countries and regions worldwide.
MGIT: A Multi-modal Global Instance Tracking Benchmark Based on Hierarchical Semantic Framework
Visual Language Tracking & Long Video Understanding and Reasoning & Hierarchical Semantic Information Annotation
This work extends the GIT task and the VideoCube benchmark by constructing a multi-modal benchmark called MGIT. The MGIT benchmark is designed to capture the complex video narrative relationships and fully encompass the intricate spatio temporal and causal connections illustrated in long videos.
MGIT is the supporting platform for research accepted by NeurIPS 2023. As of Feb. 2024, the platform has received 315k+ page views, 1k+ downloads, 400+ trackers from 130+ countries and regions worldwide.
BioDrone: A Bionic Drone-based Single Object Tracking Benchmark for Robust Vision
Visual Object Tracking & Drone-based Tracking & Robust Visual Research
BioDrone is the first bionic drone-based SOT benchmark, it features videos captured from a flapping-wing UAV system with a major camera shake due to its aerodynamics. BioDrone highlights the tracking of tiny targets with drastic changes between consecutive frames, providing a new robust vision benchmark for SOT.
BioDrone is the supporting platform for research accepted by IJCV 2023 and the organization of the 3rd High-Speed Low-Power Visual Understanding Challenge as competition data.