Text
SEGMENTASI SEMANTIK OPTIC DISC DAN OPTIC CUP PADA CITRA RETINA MENGGUNAKAN MODIFIKASI ARSITEKTUR U-NET RESIDUAL BLOCK
One of the digital image processing processes is optic disc and optic cup segmentation on the retina. The optic disc and optic cup segmentation process can use existing datasets, one of which is the MESSIDOR-2 dataset. The number of datasets and the size of the image used in the semantic segmentation process will affect the performance of a method. One method that has strong capabilities when trained with large-dimensional datasets is the Convolutional Neural Network (CNN). The main capability of CNN lies in its architecture, namely U-Net, but U�Net has a weakness in the small number of layers in it. The addition of excessive layers in the U-Net architecture will make the parameter space in the architecture enlarged, so that the U-Net architecture has difficulty in training the network, because of the many gradients that are lost in the training process. The addition of Residual Block to the U-Net architecture will make the architecture add more layers without losing the gradient by utilizing a skip connection. The semantic segmentation process carried out in this study uses a modification of the U-Net Residual Block architecture. The stages of research carried out include data collection, data preprocessing, training, testing, evaluation, as well as analysis and interpretation of the results. The results of the model performance obtained are the accuracy value of 99.79%, Intersection over Union (IoU) of 71,16%, f1 score of 81,32%, sensitivity of 77,64%, and specificity of 94,69%. It means, can be said that the U-Net Residual Block architecture is capable of segmentation optic disc and optic cup from the given image data.
Inventory Code | Barcode | Call Number | Location | Status |
---|---|---|---|---|
2207000705 | T67708 | T677082022 | Central Library (Referens) | Available but not for loan - Not for Loan |
No other version available