Semantic priors and virtual outlier synthesis enable parameter-efficient open-world object detection

Abstract Open-world object detection requires knowledge of categories and discovery of new objects never seen in training. Full model fine tuning fails often in such dynamic environment. Fully tuned models overfit known classes and lose their sensitivity to unknown objects at high computational cost...

Deskribapen osoa

Xehetasun bibliografikoak

Egile Nagusiak:	Jiaming Gu, Yehui Zheng, Yuzhou Liu, Caimei Liu, Shu Gong, Luoyang Luo
Formatua:	Artigo
Hizkuntza:	Inglês
Argitaratua:	Nature Portfolio 2026-05-01
Saila:	Scientific Reports
Gaiak:	Open-world object detection Parameter-efficient fine-tuning Virtual outlier synthesis Transformer Incremental learning Computer vision
Sarrera elektronikoa:	https://doi.org/10.1038/s41598-026-51018-8
Etiketak:	Etiketa erantsi Etiketarik gabe, Izan zaitez lehena erregistro honi etiketa jartzen!