TNet-Segmentation

• It is a pytorch based Segmentation model inspired by the research paper T. M. Khan, A. Robles-Kelly and S. S. Naqvi, "T-Net: A Resource-Constrained Tiny Convolutional Neural Network for Medical Image Segmentation," 2022 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA, 2022, pp. 1799-1808, doi: 10.1109/WACV51458.2022.00186. published in CVPR 2022

How to install

• run pip install TNet-Segmentation to install the package

How to use

Creating Model

• The model takes three parameters: -
  • input_channnel (defaults to 3)
  • emb_size (size of the output channel after first downsampling conv layer.Defaults to 512)
  • num_classes (the num of segmentation classes, defaults to 3)
• Usage is as simple as below:-

  >>> from TNet_Segmentation import TNet                                
  >>> net = TNet(input_channel=3, emb_size=256, num_classes=3) 
  

Visualizing

• After creating the model call visualize function as follows:-

  >>> from TNet_Segmentation import visualize 
  >>> visualize(net)
  =================================================================
  Layer (type:depth-idx)                   Param #
  =================================================================
  TNet                                     --
  ├─Conv2d: 1-1                            7,168
  ├─TNetConvBlock: 1-2                     --
  │    └─ModuleList: 2-1                   --
  │    │    └─Conv2d: 3-1                  590,080
  │    │    └─BatchNorm2d: 3-2             512
  │    │    └─ReLU: 3-3                    --
  │    │    └─Conv2d: 3-4                  65,792
  │    │    └─BatchNorm2d: 3-5             512
  │    │    └─Conv2d: 3-6                  2,560
  │    │    └─BatchNorm2d: 3-7             512
  │    │    └─ReLU: 3-8                    --
  │    │    └─MaxPool2d: 3-9               --
  

DiceLoss function

• Dice Loss is implemented in this library and can be used as follows:

  >>> from TNet_Segmentation import DiceLoss
  >>> loss = DiceLoss()