including Attention_UNet, Channel_UNet, FCN8s, ResNet34_UNet, UNet, UNetpp.
Window 11 + PyCharm + python3.7 + PyTorch 1.7.1
All you need to do is import the new dataset path and reference the neural network for model training and prediction.
The purpose of this repository is to perform image segmentation of the longissimus dorsi muscle of meat sheep using CT images. This repository includes the model network code (including the model training weights pth file), the model training log file and the AVER_HD, MOIU&DICE, LOSS training process diagrams. We put in 62 original CT images of longissimus dorsi muscle of meat sheep as sample files, which are in the dcm_25 folder, and 1471 manually labeled CT slices including 25 longissimus dorsi muscle of meat sheep, which are located in the label folder.In the prediction results folder£¬From left to right, the images in the prediction folder are the original CT image, the prediction image, and the mask image. In the prediction folder, we only keep the prediction results of one model, and the other five models are similar to this model. We only need to change the corresponding model to predict after network training.