This course covers the learning outcomes and goals of understanding CycleGANs, a novel deep learning tool in pathology. Participants will learn about the applications best suited for CycleGANs, their pitfalls, and how to apply them to pathology and microscopy. The course teaches skills such as model training, inference, generator and discriminator architecture, loss functions, and code implementation. The teaching method includes a code demo, model architecture explanation, and practical examples. The intended audience for this course includes individuals interested in deep learning, GANs, pathology, microscopy, and image conversion problems.
What Are CycleGANs? - A Novel Deep Learning Tool in Pathology
Overview
Syllabus
Introduction
Objectives
Colorization
Super Resolution
Unpaired Image Translation
Examples of Unpaired Problems
CycleGANs
Review
Generator and Discriminator
Generator Architecture
Discriminator Architecture
Loss Function
Code
Package
MBdev
Model Architecture
Tests
Discriminator
Identity Loss
Documentation
Transforms
Data Loader
FastAI
CycleGAN Trainer
Learning Rate Scheduling
Adding Learning Rate Scheduling to the learner object
Creating the learner class
Learning rate finder
Performing inference
Example code
Examples
Web Interface
Prediction Function
Interface
Cropping
CycleGAN failure
CycleGANs in pathology
Taught by
Abhishek Thakur
Related Courses
-
Unpaired Image to Image Translation Using CycleGAN - An Introduction
-
CycleGAN Implementation From Scratch
-
Generative Adversarial Networks and Stable Diffusion
-
Image to Image Translation Using Pix2Pix GAN
-
What Are GANs - Generative Adversarial Networks Explained
-
TransGAN - Two Transformers Can Make One Strong GAN - Machine Learning Research Paper Explained
