A new dataset and a distractor-aware architecture for transparent object tracking

Lukežič, Alan (avtor), Trojer, Žiga (avtor), Matas, Jiří (avtor), Kristan, Matej (avtor)

Performance of modern trackers degrades substantially on transparent objects compared to opaque objects. This is largely due to two distinct reasons. Transparent objects are unique in that their appearance is directly affected by the background. Furthermore, transparent object scenes often contain many visually similar objects (distractors), which often lead to tracking failure. However, development of modern tracking architectures requires large training sets, which do not exist in transparent object tracking. We present two contributions addressing the aforementioned issues. We propose the first transparent object tracking training dataset Trans2k that consists of over 2k sequences with 104,343 images overall, annotated by bounding boxes and segmentation masks. Standard trackers trained on this dataset consistently improve by up to 16%. Our second contribution is a new distractor-aware transparent object tracker (DiTra) that treats localization accuracy and target identification as separate tasks and implements them by a novel architecture. DiTra sets a new state-of-the-art in transparent object tracking and generalizes well to opaque objects.

• Jezik: Angleški jezik
• Ključne besede: visual object tracking, tracking transparent objects, deep learning, distractors
• Vrsta gradiva: Članek v reviji
• Tipologija: 1.01 - Izvirni znanstveni članek
• Organizacija: FRI - Fakulteta za računalništvo in informatiko
• Status publikacije: Objavljeno
• Različica publikacije: Objavljena publikacija
• Leto izida: 2024
• Št. strani: Str. 2729–2742
• Številčenje: Vol. 132, iss. 8
• PID: 20.500.12556/RUL-155777
• UDK: 004.93:004.8
• ISSN pri članku: 0920-5691
• DOI: 10.1007/s11263-024-02010-0
• COBISS.SI-ID: 186509827
• Datum objave v RUL: 17.04.2024

Gradivo je del revije

• Naslov: International journal of computer vision
• Skrajšan naslov: Int. j. comput. vis.
• Založnik: Springer Nature
• ISSN: 0920-5691
• COBISS.SI-ID: 25641472

Licence

• Licenca: CC BY 4.0, Creative Commons Priznanje avtorstva 4.0 Mednarodna
• Povezava: http://creativecommons.org/licenses/by/4.0/deed.sl
• Opis: To je standardna licenca Creative Commons, ki daje uporabnikom največ možnosti za nadaljnjo uporabo dela, pri čemer morajo navesti avtorja.

Sekundarni jezik

• Jezik: Slovenski jezik
• Ključne besede: vizualno sledenje objektov, sledenje prosojnih objektov, globoko učenje