1st Place Solution to ECCV-TAO-2020: Detect and Represent Any Object for Tracking
This work addresses object tracking for diverse, real-world scenarios, but it is incremental as it builds on existing paradigms and techniques.
The paper tackles the tracking-any-object problem by extending tracking-by-detection, integrating state-of-the-art detection techniques and training feature learning networks to represent objects, achieving first place in the ECCV-TAO-2020 challenge.
We extend the classical tracking-by-detection paradigm to this tracking-any-object task. Solid detection results are first extracted from TAO dataset. Some state-of-the-art techniques like \textbf{BA}lanced-\textbf{G}roup \textbf{S}oftmax (\textbf{BAGS}\cite{li2020overcoming}) and DetectoRS\cite{qiao2020detectors} are integrated during detection. Then we learned appearance features to represent any object by training feature learning networks. We ensemble several models for improving detection and feature representation. Simple linking strategies with most similar appearance features and tracklet-level post association module are finally applied to generate final tracking results. Our method is submitted as \textbf{AOA} on the challenge website. Code is available at https://github.com/feiaxyt/Winner_ECCV20_TAO.