Seite - 164 - in Joint Austrian Computer Vision and Robotics Workshop 2020

bediscriminatedbytheconcave/convexshapes. Sec- ondly, we concentrated only on description of layer 1, as imperfections in its segmentationpropagated to subsequent layers. As the segmentation algorithms mature, descriptors of remaining layers could be in- corporated. With an increased number of features, the ensemble-based detectors (FB in this work) may improve in their performance. Finally, after the seg- mentation algorithms become very advanced, it may turn out that the area-related descriptors loose their discriminative power and a need for completely new set descriptors may arise. In the proposed semi- supervised framework, the manually crafted features can be replaced by ones proposed by auto-encoders [1]orgenerativeadversarial neuralnetworks [8]. References [1] C. C. Aggarwal. Outlier analysis. In Data mining, pages75–79. Springer, 2015. [2] F. Angiulli and C. Pizzuti. Fast outlier detection in high dimensional spaces. In European Conference on Principles of Data Mining and Knowledge Dis- covery, pages 15–27. Springer, 2002. [3] C. M. Bishop. Pattern Recognition and Machine Learning, chapter 3.1.4: Regularized least squares, pages144–145. Springer, 2006. [4] M. M. Breunig, H.-P. Kriegel, R. T. Ng, and J.Sander. LOF: Identifyingdensity-based localout- liers. In ACM sigmod record, volume 29, pages 93– 104.ACM,2000. [5] M. K. Garvin, M. D. Abramoff, X. Wu, S. R. Russell, T. L. Burns, and M. Sonka. Auto- mated 3-d intraretinal layer segmentation of macu- lar spectral-domain optical coherence tomography images. IEEE Transactions on Medical Imaging, 28(9):1436–1447,Sep.2009. [6] V. Hodge. A survey of outlier detection methodolo- gies. Artificial Intelligence Review, 22:85–126, 10 2004. [7] A.LazarevicandV.Kumar. Featurebaggingforout- lier detection. In Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining, pages 157–166. ACM, 2005. [8] Y. Liu, Z. Li, C. Zhou, Y. Jiang, J. Sun, M. Wang, andX.He. Generativeadversarialactivelearningfor unsupervised outlier detection. IEEE Transactions onKnowledge andDataEngineering, 2019. [9] P.J.Mekjavic,V.J.Balciu¨niene,L.Ceklic,J.Ernest, Z.Jamrichova,Z.Z.Nagy, I.Petkova,S.Teper, I.G. Topcic, and M. Veith. The burden of macular dis- eases in central and eastern Europe —- implications forhealthcare systems. Value inHealthRegional Is- sues, 19:1–6,2019. [10] T. Otani, S. Kishi, and Y. Maruyama. Patterns of diabetic macular edema with optical coherence to- mography. American Journal of Ophthalmology, 127(6):688–693,1999. [11] Pro Visu Foundation. Fovea Centralis. https://www.provisu.ch/cgi/en/anatomical- structure.pl?en+alp+F+A09.371.729.522.436, 2018. [Online; accessed13-October-2019]. [12] P.J.RousseeuwandK.V.Driessen. Afastalgorithm for the minimum covariance determinant estimator. Technometrics, 41(3):212–223,1999. [13] B. Scho¨lkopf, J. C. Platt, J. Shawe-Taylor, A. J. Smola, and R. C. Williamson. Estimating the sup- portofahigh-dimensionaldistribution. Neuralcom- putation, 13(7):1443–1471, 2001. [14] W. D. Strain, X. Cos, and C. Pru¨nte. Considerations for management of patients with diabetic macular edema: Optimizing treatment outcomes and min- imizing safety concerns through interdisciplinary collaboration. DiabetesResearchandClinicalPrac- tice, 126:1–9,2017. 164