|
|
Publications
Shao, ZY., Yin, Y., Lyu, H.G., Sun, Sh. and Guedes Soares, C. (2024), Assessing deep learning methods for sea-surface multi object tracking from visible light video, Advances in Maritime Technology and Engineering, Guedes Soares, C. & Santos T.A. (Eds.), Taylor and Francis, London, UK, pp. 151-158.
This study assesses the effectiveness of several deep learning approaches for multi-object tracking on the sea surface, employing visible light videos captured by electro-optical sensors on unmanned surface vehicles. Utilizing the Jary Maritime Tracking Dataset, six leading deep learning methods including SORT, Deep-SORT, DeepMOT, UAVMOT, ByteTrack and BoT-SORT are quantitatively compared. Focusing on seven types of common of sea-surface objects as main interest categories for navigation, performance is quantified by calculating the cardinality of true and false positives and negatives as well as ID Switches (IDs) in the test set. Employing Multiple Object Tracking Accuracy (MOTA), Identity F1 (IDF1) score, IDs and Frames Per Second (FPS) as performance metrics, noticeable differences are found that substantial variations in the efficacy of diverse multiobject tracking methods. The results identify BoT-SORT as the most effective method regarding accuracy and stability for sea-surface object tracking, highlighting the need for refined association strategies in visual tracking
to enhance maritime navigation.
If you did not manage to obtain a copy of this paper: Request a copy of this article
For information about all CENTEC publications you can download: Download the Complete List of CENTEC Publications
|