Skip to content

Latest commit

 

History

History
381 lines (287 loc) · 14.4 KB

4_OSR.md

File metadata and controls

381 lines (287 loc) · 14.4 KB
Raw

4. Multi-Class Novelty Detection & Open Set Recognition

4.1 Classfication

[TPAMI-2013] Toward Open Set Recognition
Authors: Walter J. Scheirer, Anderson de Rezende Rocha, Archana Sapkota, Terrance E. Boult
Institution: Harvard University; University of Campinas; University of Colorado, Colorado Springs

Kicking-off paper using 1-vs-set machine for OSR.

This paper highlights the practicality of OSR by showing the difference between classification and recognition: classification only has a given set of classes between which we must discriminate; Recognition has some classes we can recognize in a much larger space of things we do not recognize. The paper shows the validity of 1-class SVM and binary SVM for OSR, and proposes 1-vs-Set SVM to manage the open-set risk by solving a two-plane optimization problem instead of the classic half-space of a binary linear classifier.

4.1.1 EVT-based Uncertainty Calibration

[TPAMI-2014] Probability models for open set recognition
Authors: Walter J. Scheirer, Lalit P. Jain, Terrance E. Boult
Institution: Harvard University; University of Colorado, Colorado Springs

W-SVM using CAP and EVT for score calibration on one-class and binary SVM.

CAP explicitly models the probability of class membership abating from ID points to OOD points, as classic probabilistic model lacks the consideration of open space, and EVT exactly focuses on modeling the tailed distribution with extreme high/low values. The novel Weibull-calibrated SVM (W-SVM) algorithm is introduced, combining the useful properties of CAP and EVT.

[ECCV-2014] Multi-class open set recognition using probability of inclusion
Authors: Lalit P. Jain, Walter J. Scheirer, Terrance E. Boult
Institution: University of Colorado, Colorado Springs; Harvard University; Securics

PI-SVM estimating the unnormalized posterior probability of class inclusion.

Modeling positive training data at the decision boundary, where we can invoke the statistical EVT. A new algorithm called the PI-SVM is introduced for estimating the unnormalized posterior probability of multiple class inclusion.

[CVPR-2016] Towards open set deep networks
Authors: Abhijit Bendale, Terrance E. Boult
Institution: University of Colorado, Colorado Springs

OpenMax: Replacing softmax layer with OpenMax and calibrating the confidence to predict novel class.

This method uses the scores from the penultimate layer to estimate if the input is “far” from known training data.

[BMVC-2017] Adversarial robustness: Softmax versus openmax
Authors: Andras Rozsa, Manuel Gunther, Terrance E. Boult
Institution: University of Colorado, Colorado Springs

4.1.2 EVT-free Calibration

[ICCV-2021] Evidential Deep Learning for Open Set Action Recognition.
Authors: Wentao Bao, Qi Yu and Yu Kong
Institution: Rochester Institute of Technology, Rochester, NY 14623, USA

[CVPR-2019] Deep transfer learning for multiple class novelty detection
Authors: Pramuditha Perera, Vishal M. Patel
Institution: Johns Hopkins University

[CVPR-2020] Generative-discriminative feature representations for open-set recognition
Authors: Pramuditha Perera, Vlad I. Morariu, Rajiv Jain, Varun Manjunatha, Curtis Wigington, Vicente Ordonez, Vishal M. Patel
Institution: Johns Hopkins University; Adobe Research; University of Virginia

[arXiv-2021] M2iosr: Maximal mutual information open set recognition
Authors: Xin Sun, Henghui Ding, Chi Zhang, Guosheng Lin, Keck-Voon Ling
Institution: Nanyang Technological University

4.1.3 Unknown Generation

[BMVC-2017] Generative openmax for multi-class open set classification
Authors: ZongYuan Ge, Sergey Demyanov, Zetao Chen, Rahil Garnavi
Institution: IBM Research; Vision for Robotics Lab

[ECCV-2018] Open set learning with counterfactual images
Authors: Lawrence Neal, Matthew Olson, Xiaoli Fern, Weng-Keen Wong, Fuxin Li;
Institution: Oregon State University

[CVPR-2021] Learning Placeholders for Open-Set Recognition
Authors: Da-Wei Zhou, Han-Jia Ye, De-Chuan Zhan
Institution: Nanjing University

[TKDE-2020] Collective decision for open set recognition
Authors: Chuanxing Geng, Songcan Chen
Institution: Nanjing University

[arXiv-2020] One-vs-rest network-based deep probabil- ity model for open set recognition
Authors: Jaeyeon Jang, Chang Ouk Kim
Institution: Yonsei University

[EUSIPCO-2019] Open-set recognition using intra-class splitting
Authors: Patrick Schlachter, Yiwen Liao, Bin Yang
Institution: University of Stuttgart

[ICCV-2021] OpenGAN: Open-Set Recognition via Open Data Generation
Authors: Shu Kong, Deva Ramanan
Institution: Carnegie Mellon University; Argo AI

4.1.4 Label Space Redesign

[Report-2021] Language guided out-of-distribution detection
Authors: William Gan
Institution: UC, Berkeley

[CVPR-2018] Hierarchical novelty detection for visual object recognition
Authors: Kibok Lee, Kimin Lee† Kyle Min, Yuting Zhang, Jinwoo Shin† Honglak Lee
Institution: University of Michigan; Korea Advanced Institute of Science and Technology; Google Brain

[CVPR-2021] Mos: Towards scaling out-of-distribution detection for large semantic space
Authors: Rui Huang, Yixuan Li
Institution: University of Wisconsin-Madison

[NeurIPS-2018] Out-of-distribution detection using multiple semantic label representations
Authors: Gabi Shalev, Yossi Adi, Joseph Keshet
Institution: Bar-Ilan University

[arXiv-2021] Learning transferable visual models from natural language supervision
Authors: Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever
Institution: OpenAI

[arXiv-2021] Exploring the limits of out-of-distribution detection
Authors: Stanislav Fort, Jie Ren, Balaji Lakshminarayanan
Institution: Stanford University; Google Research

4.2 Distance-based Method

[TPAMI-2021] Adversarial Reciprocal Points Learning for Open Set Recognition.
Authors: Guangyao Chen, Peixi Peng, Xiangqian Wang and Yonghong Tian
Institution: Peking University, Peng Cheng Laboratory

[TPAMI-2020] Convolutional Prototype Network for Open Set Recognition.
Authors: Hong-Ming Yang, Xu-Yao Zhang, Fei Yin, Qing Yang and Cheng-Lin Liu
Institution: Institute of Automation Chinese Academy of Sciences

[BMVC-2018] Metric learning for novelty and anomaly detection
Authors: Masana, Marc and Ruiz, Idoia and Serrat, Joan and van de Weijer, Joost and Lopez, Antonio M
Institution: Universitat Autonoma de Barcelona, Bellaterra, Spain

[Report] p-odn: prototype- based open deep network for open set recognition
Authors: Yu Shu, Yemin Shi, Yaowei Wang, Tiejun Huang, Yonghong Tian
Institution: Peking University; Peng Cheng Laboratory

[CVPR-2020] Few-shot open-set recognition using meta-learning
Authors: Bo Liu, Hao Kang, Haoxiang Li, Gang Hua, Nuno Vasconcelos
Institution: Wormpex AI Research; UC, San Diego

[ECCV-2020] Learning open set network with discriminative reciprocal points
Authors: Guangyao Chen, Limeng Qiao, Yemin Shi, Peixi Peng, Jia Li, Tiejun Huang, Shiliang Pu, Yonghong Tian
Institution: Peking University; Beihang University; Peng Cheng Laboratory; Hikvision Research Institute

[CVPR-2019] Classification-reconstruction learning for open-set recognition
Authors: Ryota Yoshihashi, Wen Shao, Rei Kawakami, Shaodi You2, Makoto Iida, Takeshi Naemura1
Institution: The University of Tokyo; Data61-CSIRO

[AAAI-2020] Open-set recognition with gaussian mixture variational autoencoders
Authors: Alexander Cao, Yuan Luo, Diego Klabjan
Institution: Northwestern University

[Machine Learning-2017] Nearest neighbors distance ratio open-set classifier
Authors: Pedro R. Mendes Junior, Rafael de O. Werneck, Bernardo V. Stein, Daniel V. Pazinato, Waldir R. de Almeida, Otavio A. B. Penatti, Ricardo da S. Torres, Anderson Rocha, Roberto M. de Souza, Otavio A. B. Penatti
Institution: University of Campinas; SAMSUNG Research Institute

4.3 Reconstruction

4.3.1 Sparse Representation

[TPAMI-2016] Sparse representation-based open set recognition
Authors: He Zhang, Vishal M. Patel
Institution: Rutgers University

SROSR models the tails of the matched and sum of non-matched reconstruction error distributions.

This method model the tail of the above two error distributions using the statistical EVT, and the confidence scores corresponding to the tail distributions of a novel test sample are then fused to determine its identity. Notice that the hidden embedding during reconstruction is regularized by sparsity.

[CVPR-2013] Kernel null space methods for novelty detection
Authors: Paul Bodesheim, Alexander Freytag, Erik Rodner, Michael Kemmler, Joachim Denzler
Institution: University Jena; UC Berkeley

[CVPR-2017] Incremental kernel null space discriminant analysis for novelty detection
Authors: Juncheng Liu, Zhouhui Lian, Yi Wang, Jianguo Xiao
Institution: Peking University; Dalian University

4.3.2 Reconstruction-Error

[CVPR-2019] C2AE: Class Conditioned Auto-Encoder for Open-Set Recognition
Authors: Poojan Oza, Vishal M. Patel
Institution: Johns Hopkins University

[CVPR-2020] Conditional gaussian distribution learning for open set recognition
Authors: Xin Sun, Zhenning Yang, Chi Zhang, Keck-Voon Ling, Guohao Peng
Institution: Nanyang Technological University

[CVPR-2021] Counterfactual zero-shot and open-set visual recognition
Authors: Zhongqi Yue, Tan Wang, Qianru Sun, Xian-Sheng Hua, Hanwang Zhang
Institution: Nanyang Technological University; Singapore Management University; Alibaba Damo Academy

[ECCV-2020] Open-set adversarial defense
Authors: Rui Shao, Pramuditha Perera, Pong C. Yuen, Vishal M. Patel
Institution: Hong Kong Baptist University; AWS AI Labs; Johns Hopkins University