3D Segmentations models for pytorch

Supported architectures:

• Unet
• Unet++

Install:

git clone git@github.com:EIDOSLAB/torch-segmentation-3d.git && cd torch-segmentation-3d
pip3 install .

Example:

import torch_segmentation_3d as seg3d
model = seg3d.models.Unet(encoder_name="resnet18", encoder_weights="imagenet", in_channels=1, classes=3)

Supported encoders:

• ResNet

  Encoder	Weights	Params
  resnet18	imagenet	33M
  resnet34	imagenet	63M
  resnet50	imagenet	46M
  resnet101	imagenet	85M

Cache location:

• Linux: ~/.cache/torch/hub/checkpoints/resnet18-78569d5e.pth (imagenet-1k)