Naslov
Automatic universal taxonomies for multi-domain semantic segmentation

Autori
Bevandić, Petra ; Šegvić, Siniša

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Proceedings of the 33rd British Machine Vision Conference / - London : Delhi : BMVA Press, 2022, 1-10

Skup
British Machine Vision Conference (BMVC)

Mjesto i datum
London, Ujedinjeno Kraljevstvo, 21.11.2022. - 24.11.2022

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
Semantic segmentation, universal taxonomies

Sažetak
Training semantic segmentation models on multiple datasets has sparked a lot of recent interest in the computer vision community. This interest has been motivated by expensive annotations and a desire to achieve proficiency across multiple visual domains. However, established datasets have mutually incompatible labels which disrupt principled inference in the wild. We address this issue by automatic construction of universal taxonomies through iterative dataset integration. Our method detects subset-superset relationships between dataset-specific labels, and supports learning of sub-class logits by treating super-classes as partial labels. We present experiments on collections of standard datasets and demonstrate competitive generalization performance with respect to previous work.

Izvorni jezik
Engleski

Znanstvena područja
Računarstvo

POVEZANOST RADA