| Title |
Research of retinal vessels segmentation by fully convolutional networks / |
| Translation of Title |
Tinklainės kraujagyslių segmentavimo tyrimas taikant konvoliucinius neuroninius tinklus. |
| Authors |
Mune Gowda, Divyesh |
| Full Text |
|
| Pages |
65 |
| Keywords [eng] |
image segmentation ; retinal vessels segmentation ; unet ; fully convolutional networks |
| Abstract [eng] |
In this research, the retinal vessel segmentation by a fully convolutional neural network is studied in detail. Retinal vessel segmentation is helpful in the diagnosis of diabetic retinopathy, hypertension and arteriosclerosis. This research aims to improve convolutional neural networks to provide efficient retinal vessel segmentation results. The UNET was introduced in 2015, which provided better efficient segmentation for biomedical images with a fewer dataset. In this research, the UNET and possible modified and improved networks based on UNET was built and tested for efficient segmentation of retinal vessels. A fully convolution network UNET G UCDA is proposed in this project which was found during the research to provide efficient retinal vessel segmentation compared to all other UNET based networks that were trained and evaluated in this research. |
| Dissertation Institution |
Kauno technologijos universitetas. |
| Type |
Master thesis |
| Language |
English |
| Publication date |
2021 |
