Difficulty-Aware Simulator for Open Set Recognition

Moon, WonJun; Park, Junho; Seong, Hyun Seok; Cho, Cheol-Ho; Heo, Jae-Pil

doi:10.1007/978-3-031-19806-9_21

Cited by 29 publications

(14 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Apart from the above baselines, other OSR methods are not considered for a fair comparison, since they require additional parameters to be optimized in the encoder-decoder structure [34,43,9,50] or the generator-discriminator structure [19,33,32].…”

Section: Baseline Methodsmentioning

confidence: 99%

R-Tuning: Regularized Prompt Tuning in Open-Set Scenarios

Li¹,

Zhang²,

Cao³

et al. 2023

Preprint

View full text Add to dashboard Cite

In realistic open-set scenarios where labels of a part of testing data are totally unknown, current prompt methods on vision-language (VL) models always predict the unknown classes as the downstream training classes. The exhibited label bias causes difficulty in the open set recognition (OSR), by which an image should be correctly predicted as one of the known classes or the unknown one. To learn prompts in open-set scenarios, we propose the Regularized prompt Tuning (R-Tuning) to mitigate the label bias. It introduces open words from the WordNet to extend the range of words forming the prompt texts from only closedset label words to more. Thus, prompts are tuned in a simulated open-set scenario. Besides, inspired by the observation that classifying directly on large datasets causes a much higher false positive rate than on small datasets, we propose the Combinatorial Tuning and Testing (CTT) strategy for improving performance. CTT decomposes R-Tuning on large datasets as multiple independent group-wise tuning on fewer classes, then makes comprehensive predictions by selecting the optimal sub-prompt. For fair comparisons, we construct new baselines for OSR based on VL models, especially for prompt methods. Our method achieves the best results on datasets with various scales. Extensive ablation studies validate the effectiveness of our method.

show abstract

Section: Baseline Methodsmentioning

confidence: 99%

R-Tuning: Regularized Prompt Tuning in Open-Set Scenarios

Li¹,

Zhang²,

Cao³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Numerous researchers believe that modeling only known classes is insufficient and suggest that incorporating prior knowledge about unknown classes is necessary. Some approaches attempt to generate fake data [30], counterfactual images [31] or confused samples [17], [18]. Others [32], [33] introduce background classes or known unknown classes (KUC) to represent the unknown.…”

Section: B Open Set Recognitionmentioning

confidence: 99%

“…We compared our method against other state-of-the-art open set recognition approaches, including Softmax, Open-Max [25], ARPL [17], AKPF [18], Objectosphere Loss [32] and DIAS [30]. Accordingly, OpenMax calibrates the prediction probability by Weibull distribution.…”

Section: Comparison With the State-of-the-artsmentioning

confidence: 99%

Towards open-set text recognition via label-to-prototype learning

Liu

Yang²,

Qin³

et al. 2023

Pattern Recognition

View full text Add to dashboard Cite

“…To achieve this, we leverage the different layers in a network to generate samples of different difficulty levels, so that they can scatter both inside and outside of the known-classes distribution. Inspired by DiAS [18], we complement an architecture called Synthetic Unknown Attack Sample Generator (SUASG) to produce synthetic unknown samples for FAS during training (upper-left of Fig. 1).…”

Section: Domain Generalized Unknown Attacksmentioning

confidence: 99%

“…On the other hand, as the simulated data are synthesized from the original training data, the original labels could serve as the pseudo labels for the synthesized data. We thus smooth them to form the target probabilities of the synthetic data as suggested by [18]:…”

Section: Dataset Number Of Videos/imagesmentioning

confidence: 99%

Domain-Generalized Face Anti-Spoofing with Unknown Attacks

Hong,

Lin,

Liu

et al. 2023

2023 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Although face anti-spoofing (FAS) methods have achieved remarkable performance on specific domains or attack types, few studies have focused on the simultaneous presence of domain changes and unknown attacks, which is closer to real application scenarios. To handle domain-generalized unknown attacks, we introduce a new method, DGUA-FAS 1 , which consists of a Transformer-based feature extractor and a synthetic unknown attack sample generator (SUASG). The SUASG network simulates unknown attack samples to assist the training of the feature extractor. Experimental results show that our method achieves superior performance on domain generalization FAS with known or unknown attacks.

show abstract

Difficulty-Aware Simulator for Open Set Recognition

Cited by 29 publications

References 34 publications

R-Tuning: Regularized Prompt Tuning in Open-Set Scenarios

R-Tuning: Regularized Prompt Tuning in Open-Set Scenarios

Towards open-set text recognition via label-to-prototype learning

Domain-Generalized Face Anti-Spoofing with Unknown Attacks

Contact Info

Product

Resources

About