Tantárgy adatlapja
Tárgy neve: Open Set Recognition Methodologies
Tárgy kódja: P-DO_0287
Óraszám: N: 2/2/0, L: 0/0/0
Kreditérték: 6
Az oktatás nyelve: angol
Követelmény típus: Gyakorlati jegy
Felelős kar: ITK
Felelős szervezeti egység: ITK Doktori és Habilitációs Iroda
Tárgyfelelős oktató:
Dr. Tornai Kálmán
Tárgyleírás:
The objective of the course is to review the fundamentals of i) the open set recognition problem and its mathematical formulation, ii) basic approaches for open set recognition, iii) recent solutions.
Discussed topics are
- Closed-set vs open-set problems (classification and recognition)
- Mathematical formulation, open-set risk
- Anomaly and novelty detection
- Out-of-distribution detection
- Open-set recognition approaches
- SVM based solutions
- Extreme Value Theory
- Extreme Value Machine
- Deep Set Networks (OpenMax)
- Generative Approaches
- Datasets for evaluation
- Evaluation metrics and protocols
List of selected literature:
- M. Salehi, H. Mirzaei, D. Hendrycks, Y. Li, M. Rohban, and M. Sabokrou: A Unified Survey on Anomaly, Novelty, Open-Set, and Out-of-Distribution Detection: Solutions and Future Challenges, https://doi.org/10.48550/arXiv.2110.14051
- Walter J. Scheirer and Anderson Rocha and Archana Sapkota and Terrance E. Boult: Towards Open Set Recognition, IEEE Transactions on Pattern Analysis and Machine Learning, 2013, 35 (7)
- Ethan Rudd, Lalit P. Jain, Walter J. Scheirer, Terrance Boult, The Extreme Value Machine, IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2018, 40 (3)
- Ethan Rudd, Lalit P. Jain, Walter J. Scheirer, Terrance Boult, Towards Open Set Deep Networks, Ethan Rudd, Lalit P. Jain, Walter J. Scheirer, Terrance Boult,
