2021-06-022021-06-022020-03-22Civit Masot, J., Muñoz Saavedra, L., Luna Perejón, F., Montes-Sánchez, J.M. y Domínguez Morales, M.J. (2020). Incremental Learning For Fundus Image Segmentation. En eTELEMED 2020: The Twelfth International Conference on eHealth, Telemedicine, and Social Medicine (5-8), Valencia: IARIA XPS Press.978-1-61208-763-42308-4359https://hdl.handle.net/11441/111305Automated Fundus image segmentation is tradition-ally done in the image acquisition instrument and, thus, in thiscase it only needs to be able to segment data from this acquisitionsource. Cloud providers support multi GPU and TPU virtualmachines making attractive the idea of cloud-based segmentationan interesting possibility. To implement this idea we need to makecorrect predictions for fundus coming from different sources.In this paper we study the possibility of building a web basesegmentation service using incremental training, i.e, we initiallytrain the system using data from a single data set and, afterwards,perform retraining with data from other acquisition sources. Weare able to show that this type of training is efficient and canprovide good results suitable for web-based segmentation.application/pdf3engDeep learningIncremental LearningU-NetImage SegmentationEye FundusOptic discGlaucoma DetectionIncremental Learning For Fundus Image Segmentationinfo:eu-repo/semantics/conferenceObjectinfo:eu-repo/semantics/openAccess