Optic Disc and Vessel Segmentation from Fundus Image using Miniunet Architecture

The segmentation of the retinal blood vessels significantly impacts the early diagnosis of conditions that affect the eyes, such as diabetes and glaucoma. This work uses fundus pictures to segment retinal blood vessels using mini-Unet algorithms. Many thousands of gloss training samples are required for effective deep network training, which is highly permissible. Mini-Unet starts with a spatial concentration element that multiplies the concentration map by the input feature for adaptive element enhancement and deduces the concentration map in addition to the spatial measurement. Mini-unit architecture creates a structure of a constricting path that enables a particular position. We design three types of primal action, such as Unet_Gatting, AttnGating Block, and convolution, for image segmentation. The mini_unet model is a modified version of the unet model. In Figure 1.1, mention where to modify my techniques. The success of the planned network structure was verified by two segmentation responsibilities: retina vessel segmentation and lung segmentation. Our technique used segmentation of existing DRIVE and STARE datasets. In fundus images using deep learning-based convolutional neural networks. In our technique, accuracy is .95337.