Optimized High Resolution 3D Dense-U-Net Network for Brain and Spine Segmentation
KOLAŘÍK, M. BURGET, R. UHER, V. ŘÍHA, K.
A complete open-source code to Paper "Optimized High Resolution 3D Dense-U-Net Network for Brain and Spine Segmentation" published at MDPI Applied sciences journal - https://www.mdpi.com/2076-3417/9/3/404. Most useful parts of this software are python keras scripts with source code for 2D and 3D Dense-U-Net network models. Its an upgraded U-Net that we obtained better results than with classic U-Net and current hardware is capable of fitting it into GPU memory. 3D and 2D versions of Dense-U-Net, Res-U-Net and classic U-Net implementations are included.
brain; keras; segmentation; spine; u-net
Kancelář SE5.123 (Kolařík)