Collective Decision for Open Set Recognition

Geng, Chuanxing; Chen, Songcan

doi:10.1109/tkde.2020.2978199

Cited by 36 publications

(28 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Note that [21] does not explicitly give the relationships among C TA , C TR , and C TE . In most existing works [22], [61]- [63], the relationship, C TA = C TR ⊆ C TE , holds by default. Besides, the authors in [64] specifically give the following relationship: C TA ⊆ C TR ⊆ C TE , which contains the former case.…”

Section: Basic Notation and Related Definitionmentioning

confidence: 99%

“…Next, we first give a review from the discriminative model perspective, where most existing OSR algorithms are modeled from this perspective. Deep Neural Network-based [25], [64], [79]- [87] Generative model Instance Generation-based [62], [88]- [91] Non-Instance Generation-based [63] A. Discriminative Model for OSR 1) Traditional ML Methods-based OSR Models: As mentioned above, traditional machine learning methods (e.g., SVM, sparse representation, Nearest Neighbor, etc.) usually assume that the training and testing data are drawn from the same distribution.…”

Section: A Categorization Of Osr Techniquesmentioning

confidence: 99%

“…In addition, other metrics could also be used to replace Euclidean metric, and even the feature space considered could be a transformed one, as suggested by the authors. One limitation of OSNN is that just selecting two reference samples coming from different classes for comparison makes OSNN vulnerable to outliers [63].…”

Section: Instance Generationmentioning

confidence: 99%

“…Thus the threshold plays a key role. However, at the moment, the selection for it usually depends on the knowledge from KKCs, which inevitably incurs risks due to lacking available information from UUCs [63]. In fact, as the KUCs' data is often available at hand [25], [114], [115], we can fully leverage them to reduce such a risk and further improve the robustness of the OSR models for UUCs.…”

Section: Instance Generationmentioning

confidence: 99%

“…With slight modification to hierarchical Dirichlet process (HDP), Geng and Chen [63] adapted HDP to OSR and proposed the collective decision-based OSR model (CD-OSR), which can address both batch and individual samples. CD-OSR first performs a co-clustering process to obtain the appropriate parameters in the training phase.…”

Section: B Generative Model For Open Set Recognitionmentioning

confidence: 99%

See 4 more Smart Citations

Recent Advances in Open Set Recognition: A Survey

Geng

Huang

Chen

2021

IEEE Trans. Pattern Anal. Mach. Intell.

Self Cite

586

268

View full text Add to dashboard Cite

In real-world recognition/classification tasks, limited by various objective factors, it is usually difficult to collect training samples to exhaust all classes when training a recognizer or classifier. A more realistic scenario is open set recognition (OSR), where incomplete knowledge of the world exists at training time, and unknown classes can be submitted to an algorithm during testing, requiring the classifiers to not only accurately classify the seen classes, but also effectively deal with the unseen ones. This paper provides a comprehensive survey of existing open set recognition techniques covering various aspects ranging from related definitions, representations of models, datasets, evaluation criteria, and algorithm comparisons. Furthermore, we briefly analyze the relationships between OSR and its related tasks including zero-shot, one-shot (few-shot) recognition/learning techniques, classification with reject option, and so forth. Additionally, we also overview the open world recognition which can be seen as a natural extension of OSR. Importantly, we highlight the limitations of existing approaches and point out some promising subsequent research directions in this field.

show abstract