/Eff-UNet

U-Net network architecture for semantic segmentation with EfficientNet-B0 encoder

Primary LanguagePythonMIT LicenseMIT

Eff-UNet

Implementation of U-Net network architecture for semantic segmentation with EfficientNet-B0 encoder using PyTorch.

Eff-UNet

Model:

  • Total params: 5,831,968
  • Trainable params: 5,831,968
  • Non-trainable params: 0

Prerequisites

  • Python 3
  • CPU or NVIDIA GPU + CuDNN
  • Dependencies:
    • torch >= 1.4.0
    • torchvision >= 0.5.0
    • numpy >= 1.18.1
    • opencv-python >= 4.4.0.44
    • catalyst >= 20.3.3
    • albumentations >= 0.4.5
    • comet-ml >= 3.1.3