U-Net semantic image segmentation combat: training their own data sets

U-Net is a semantic image segmentation method based on the depth of learning, in particular, excellent in medical image segmentation.

This course will teach you to use the hand labelme image annotation tool to create their own data sets generated Mask image, and use the U-Net training their own data set, so that it can carry out its own image segmentation application.

Course Link: https://edu.51cto.com/course/18936.html

The course has three projects practice:

(. 1) salt Kaggle recognition match: using U-Net for recognition salt Kaggle

(2) Pothole semantic segmentation: vehicle driving road scene pits label and semantic segmentation

(. 3) Kaggle nuclear segmentation matches: U-Net for using nuclear division Kaggle

This course uses keras version of the U-Net, on the Ubuntu system demonstration project to do with Jupyter Notebook. Comprising: tagging data set, and data format conversion set Mask image generation program file written U-Net, their training data set, testing the trained network model, the performance evaluation.

This course provides project data sets and Python program files.

Course Example 1: U-Net for recognition salt Kaggle

Course Example 2: U-Net for semantic segmentation Pothole

Course Example 3: U-Net were Kaggle nuclear division