Transductive Zero-Shot Recognition via Shared Model Space Learning

Benchmark (surveying)
DOI: 10.1609/aaai.v30i1.10448 Publication Date: 2022-06-23T23:26:27Z
ABSTRACT
Zero-shot Recognition (ZSR) is to learn recognition models for novel classes without labeled data. It a challenging task and has drawn considerable attention in recent years. The basic idea transfer knowledge from seen via the shared attributes. This paper focus on transductive ZSR, i.e., we have unlabeled data classes. Instead of learning separately as existing works, put forward joint approach which learns model space (SMS) such that can be effectively transferred between using An effective algorithm proposed optimization. We conduct comprehensive experiments three benchmark datasets ZSR. results demonstrates SMS significantly outperform state-of-the-art related approaches validates its efficacy ZSR task.
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (0)
CITATIONS (32)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....