Icon

01_​Train GAN for Image Generation

01_Train GAN for Image Generation
Preprocessing- Randomly sample dataset- Efficient image load with Parallel Loop- Resize to desired size (32x32, 64x64...).NOTE: the size must match thediscriminator model input and thegenerator model output- Convert to Float and Normalize to -1,1range GAN Models 128x128x3NOTE: Generator output and discriminator inputmust match the size of the images Training GAN- Set training parameters- Select a folder to save intermediateresults.After training the generator model isprovided in output. GAN Models 64x64x3Alternative models for smaller images (lessparameters therefore reduced training time) Model executionUse the trained generator model toproduce images. Select "To Image (Auto-mapping)" conversion for the last layeroutput PostprocessingFit images to correct representation- Swap dimensions to X,Y,Channel- Denormalize and convert to unsignedbyte type Export model VisualizationShow generated images in a Tile View Training GAN for image generationRead and preprocess an image dataset to train a GAN (Generative Adversarial Network) that produces similar images.The Discriminator and Generator models must match image dimension. You can use the provided models or definenew ones. Read image foldermodelgenerator_128GANInputs:- Discriminator model- Generator model- Images in KNIPImage formatOutputs a table of:- path- imagesmodeldiscriminator_128GANmodelgenerator_64GANmodeldiscriminator_64GANRandom vectors from latent spaceSave generator model List Files/Folders Keras NetworkExecutor DL PythonNetwork Creator GAN Learner Load and preprocessimages for GAN Renderer to Image DL PythonNetwork Creator Tile View DL PythonNetwork Creator DL PythonNetwork Creator Create LantentSpace Vectors Keras NetworkWriter PostprocessGAN Images Preprocessing- Randomly sample dataset- Efficient image load with Parallel Loop- Resize to desired size (32x32, 64x64...).NOTE: the size must match thediscriminator model input and thegenerator model output- Convert to Float and Normalize to -1,1range GAN Models 128x128x3NOTE: Generator output and discriminator inputmust match the size of the images Training GAN- Set training parameters- Select a folder to save intermediateresults.After training the generator model isprovided in output. GAN Models 64x64x3Alternative models for smaller images (lessparameters therefore reduced training time) Model executionUse the trained generator model toproduce images. Select "To Image (Auto-mapping)" conversion for the last layeroutput PostprocessingFit images to correct representation- Swap dimensions to X,Y,Channel- Denormalize and convert to unsignedbyte type Export model VisualizationShow generated images in a Tile View Training GAN for image generationRead and preprocess an image dataset to train a GAN (Generative Adversarial Network) that produces similar images.The Discriminator and Generator models must match image dimension. You can use the provided models or definenew ones. Read image foldermodelgenerator_128GANInputs:- Discriminator model- Generator model- Images in KNIPImage formatOutputs a table of:- path- imagesmodeldiscriminator_128GANmodelgenerator_64GANmodeldiscriminator_64GANRandom vectors from latent spaceSave generator model List Files/Folders Keras NetworkExecutor DL PythonNetwork Creator GAN Learner Load and preprocessimages for GAN Renderer to Image DL PythonNetwork Creator Tile View DL PythonNetwork Creator DL PythonNetwork Creator Create LantentSpace Vectors Keras NetworkWriter PostprocessGAN Images

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: emilio.silvestri
Created at: 2021-07-21
On NodePit since: 2022-09-16
Last update: 2024-12-21
Created with KNIME version: v4.6.1
Tags: deep learningGANimage generation

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!