Clara Train SDK
v2.0

User Guide

  • Clara Training Framework
    • Overview
    • Installation
    • Key New Features
      • Automatic mixed precision
      • Determinism
      • Smart Cache
      • Federated learning
      • New loss functions and models
        • Loss functions
        • Models
    • Medical Model Archive
    • Bring your own components
    • Working with classification and segmentation models
    • Appendix
  • Clara Train API
  • AI-Assisted Annotation
Clara Train SDK
  • Docs »
  • Clara Training Framework »
  • Key new features »
  • New loss functions and models

New loss functions and models¶

The following list of losses and models have been newly added for this release:

Loss functions¶

  • Cross entropy

  • Dice

  • Hybrid loss between cross entropy and dice loss

  • Focal loss

  • Hybrid loss between focal loss and dice loss

Models¶

  • AlexNet for 2D/3D images on image classification with/without (fixed input size) multiple instance learning (MIL)

  • DenseNet 121, DenseNet169, DenseNet 201 for 2D/3D Classification

  • Res-UNet for 3D segmentation

  • UNet for 2D/3D segmentation

  • AHNet for 2D/3D segmentation

Next Previous

© Copyright 2019, NVIDIA

Built with Sphinx using a theme provided by Read the Docs.