[OpenR8 solution] Image-Segmentation-UNet-Keras (Using UNet and Keras framework for Pill)
infer.png
 
This solution uses Keras UNet to learn. Most of the research on medical image segmentation is based on UNet. This is a split network consisting of two parts, feature extraction and upsampling. Since the network structure is U-shaped, it is called UNet.

 

This network can use a small amount of data for training, get more accurate segmentation results, and apply to large number of segmentation problems.

 

First, we need to prepare images for the model to learn, and divide the images into different folders according to training and testing. Then use labelme to complete the label, and finally through the OpenR8 process to complete the training and inference.
 

Recommended Article

1.
OpenR8 AI Server - AI Software for Everyone (Download)