Icon

00_​Keras_​Transfer_​Learning

Read Images and Train VGG
Train ModelThis workflow reads image patches downloaded and prepared by theprevious workflows in the workflow group. It loads the VGG16 model, trainsand fine tunes the output layers. Predictions are made on the hold-out set ofimages. Read VGG16 model and add flatten / add layers Read in training data and prep for VGG16 Model Download Dataset This workflow allows the user to download the dataset. Note that the tar.gz file is ~ 2Gb.Original research article:Meng, Tao, et al. "Histology image classification using supervised classification and multimodalfusion." Multimedia (ISM), 2010 IEEE International Symposium on. IEEE, 2010.Dataset information available at:https://ome.grc.nia.nih.gov/iicbu2008/lymphoma/index.htmlDownload file from:https://ome.grc.nia.nih.gov/iicbu2008/lymphoma.tar.gz Preprocessing The workflow reads the downloaded image files and partitions into a trainingand hold-out set. Images are then split into multiple 64x64px patches thatcan be used by the VGG16 network in the following model. Histopathology: Classifying Cancer Cells with KerasWhen running this workflow completely finish executing each section before starting the next. Each portion of theworkflow creates files referenced by the next.1) Download the Dataset: this section automatically downloads the required images for you!2) Preprocessing: this section reads all the images into a knime format and chops them up into patches that will be fedinto our model later3) Train Model: the final section reads VGG16, adds layers, trains, and scores our new model. Download Histology ImagesRead VGG16Define the data directorywhere the data will be downloaded.Unzip the fileCache linksAppendclass columnReading testing patchesGroup by imageDetermine maxclassOnly train 3 new layersFlatten output to 2048 neurons64 neuronsReLUDrop Rate = 0.53 neuronsSoftmaxrenameArrange classes into [3] array for input intoNetworkReading testing patchestestPartition ImagestrainRead imagelocationsRandomlydownsample datato run workflow morequicky.Remove to train and teston full data set. GET Request Binary Objectsto Files Keras NetworkReader Create Directory Table Rowto Variable Unzip Files Table Writer List Files Rule Engine Joiner Table Reader Scorer GroupBy Many to One Keras Freeze Layers Keras Flatten Layer Keras Dense Layer Keras Dropout Layer Keras Dense Layer Keras NetworkLearner Column Rename One to Many Table Reader Table Writer Category To Number Load and preprocessimages (Local Files) Partitioning Table Writer Load and preprocessimages (Local Files) Table Reader Row Sampling Keras NetworkExecutor Train ModelThis workflow reads image patches downloaded and prepared by theprevious workflows in the workflow group. It loads the VGG16 model, trainsand fine tunes the output layers. Predictions are made on the hold-out set ofimages. Read VGG16 model and add flatten / add layers Read in training data and prep for VGG16 Model Download Dataset This workflow allows the user to download the dataset. Note that the tar.gz file is ~ 2Gb.Original research article:Meng, Tao, et al. "Histology image classification using supervised classification and multimodalfusion." Multimedia (ISM), 2010 IEEE International Symposium on. IEEE, 2010.Dataset information available at:https://ome.grc.nia.nih.gov/iicbu2008/lymphoma/index.htmlDownload file from:https://ome.grc.nia.nih.gov/iicbu2008/lymphoma.tar.gz Preprocessing The workflow reads the downloaded image files and partitions into a trainingand hold-out set. Images are then split into multiple 64x64px patches thatcan be used by the VGG16 network in the following model. Histopathology: Classifying Cancer Cells with KerasWhen running this workflow completely finish executing each section before starting the next. Each portion of theworkflow creates files referenced by the next.1) Download the Dataset: this section automatically downloads the required images for you!2) Preprocessing: this section reads all the images into a knime format and chops them up into patches that will be fedinto our model later3) Train Model: the final section reads VGG16, adds layers, trains, and scores our new model. Download Histology ImagesRead VGG16Define the data directorywhere the data will be downloaded.Unzip the fileCache linksAppendclass columnReading testing patchesGroup by imageDetermine maxclassOnly train 3 new layersFlatten output to 2048 neurons64 neuronsReLUDrop Rate = 0.53 neuronsSoftmaxrenameArrange classes into [3] array for input intoNetworkReading testing patchestestPartition ImagestrainRead imagelocationsRandomlydownsample datato run workflow morequicky.Remove to train and teston full data set. GET Request Binary Objectsto Files Keras NetworkReader Create Directory Table Rowto Variable Unzip Files Table Writer List Files Rule Engine Joiner Table Reader Scorer GroupBy Many to One Keras Freeze Layers Keras Flatten Layer Keras Dense Layer Keras Dropout Layer Keras Dense Layer Keras NetworkLearner Column Rename One to Many Table Reader Table Writer Category To Number Load and preprocessimages (Local Files) Partitioning Table Writer Load and preprocessimages (Local Files) Table Reader Row Sampling Keras NetworkExecutor

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: jon
Created at: 2017-10-28
On NodePit since: 2019-08-20
Last update: 2024-04-18
Created with KNIME version: v4.0.1
Tags: Image AnalysisDeep LearningKerasLife sciences

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

Try NodePit Runner!