Deep Learning

Deep learning-based classification is the most powerful and widely used classification method in the Visiopharm toolbox. It often significantly outperforms other classification methods for both standard and complex analysis tasks. It involves neural network algorithms that use many layers of nonlinear processing units for feature extraction and transformation with each successive layer using the output from the previous layer as input. This model structure makes deep learning among the more advanced classification methods available. Compared to other approaches, it introduces additional concepts, workflows, and configuration options.

This section aims to demystify deep learning by introducing the fundamental concepts behind deep neural networks, explaining how deep learning models are used within Visiopharm, and documenting the available advanced configuration options.

The Deep Learning documentation is divided into three main sections:

Deep Learning Models and Network Architectures

This section provides a general introduction to deep learning concepts. It explains, at a conceptual level, how deep learning models are built, how they learn from data, and how they generate predictions. It also introduces the different neural network architectures supported in Visiopharm, U-Net (Default), DeepLabv3+ and FCN-8s, and explains how they differ from each other.

Using Deep Learning in Visiopharm

This section focuses on how to use deep learning for classification in Visiopharm. It provides a step-by-step guide to creating, training, and applying a deep learning app in Visiopharm. It also explains how to understand and configure the basic training parameters, how to handle common errors, as well as how to use TensorBoard to monitor and better understand the training process.

AI Architect - Advanced Deep Learning Settings

This section explains how to configure and fine-tune deep learning models using the AI Architect in order to optimize performance for specific tasks and datasets.