Icon

Image_​dynamism_​classifier


The “Image Dynamism Classifier" workflow has three important parts. 1. The red surrounded area includes the deep learning model developed with ResNet50 and using more than 8,000 annotated images. 2. The blue area allows the workflow to use python packages through the “Conda Environment Propagation” node for predicting the classification of new images using the “Keras Network executor”. The last 3 nodes in the blue area allow column renaming and saving the results in an excel file on the user’s local machine. 3. The three yellow annotations allow the user to use a sample of pictures that we provide, a list of pictures from a user folder, or a column with a list of URLS. Each of these can be connected to the Keras Network Executor for Analysis.

This is an evolving product and so far reached an average accuracy of 81.4% on a 10-fold cross validated sample.


3) The "excel reader" node below reads imagescoming from a table having a column with URLs. Inthis case tweets from @cocacolaco. 2) The "list file/folder" node uses an internal folder of the workflowwith 10 random images. Users can read from "Local" to read afolder in their machine. This workflow reads and predicts the probability of dynamism (ongoing action) in an image. Images need to be read and preprocessed beforeusing/connecting them in the classifier (Keras Network Executor). In file 1), images are already read and pre-processed. The last part of theworkflow allow users to save the classified (static vs. dynamic) images in their machine 1) Sample of testing images used toassess the accuracy of classifier(details on article) read master.model Downloads required Conda PackagesAssign predicted class valueCreate"Prediction" columnApply model to any of these: 1) Sample Testing Set2) User's folder3) Column with URLsExport table to a folder in userlocal machineNormalize and ResizeNormalize and ResizeColumn URLsUser's FolderSample of Testing SetModel to predict dynamismConda EnvironmentPropagation Rule Engine Many to One Keras NetworkExecutor Excel Writer Pre-Processing Pre-Processing Excel Reader List Files/Folders Table Reader Model Reader 3) The "excel reader" node below reads imagescoming from a table having a column with URLs. Inthis case tweets from @cocacolaco. 2) The "list file/folder" node uses an internal folder of the workflowwith 10 random images. Users can read from "Local" to read afolder in their machine. This workflow reads and predicts the probability of dynamism (ongoing action) in an image. Images need to be read and preprocessed beforeusing/connecting them in the classifier (Keras Network Executor). In file 1), images are already read and pre-processed. The last part of theworkflow allow users to save the classified (static vs. dynamic) images in their machine 1) Sample of testing images used toassess the accuracy of classifier(details on article) read master.model Downloads required Conda PackagesAssign predicted class valueCreate"Prediction" columnApply model to any of these: 1) Sample Testing Set2) User's folder3) Column with URLsExport table to a folder in userlocal machineNormalize and ResizeNormalize and ResizeColumn URLsUser's FolderSample of Testing SetModel to predict dynamismConda EnvironmentPropagation Rule Engine Many to One Keras NetworkExecutor Excel Writer Pre-Processing Pre-Processing Excel Reader List Files/Folders Table Reader Model Reader

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: fvill
Created at: 2021-01-26
On NodePit since: 2023-03-08
Last update: 2024-05-01
Created with KNIME version: v4.7.4
Tags: marketingdynamismdeep learning

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

Try NodePit Runner!