AI art log automator

Abstract

The invention provides an automated way to categorise images using Machine Learning (ML) / Artificial Intelligence (AI) methods and outputs the results in a structured data format. The invention identif

Claims

1. A Machine Learning / Artificial Intelligence method that categorises images using a trained neural network-based algorithm, which is characterised in that it includes:Performs preprocessing on raw image data to both preformat and assign the correct metadata categorisation labels, which are comprised of user-provided categories;Classifies the image features to obtain a category label and corresponding accuracy;The tool presents a summary document to the end user that provides a list of the reviewed images, their assigned categorisation labels, and the accuracy of assignment in a predefined format.

2. The image processing method according to claim 1, wherein the feature extraction processing is performed on reformatted raw images to obtain the image features relevant to the provided categorisations:Performs dimension reduction processing on the raw image data through the dimension reduction layer of the machine learning model to obtain a reduced array image;Samples the dimensionality reduction picture through an intermediate sampling layer of the machine learning model to obtain a feature map;Transforms the feature map through the adaptive pooling layer of the machine learning model to obtain a feature vector corresponding to the feature map;Combines at least two of the feature vectors to obtain an image feature to determine a category assignment.

3. The image preprocessing method according to claim 1, wherein the raw images provided for both the model training and inference are reduced to a standardised dimensionality and format:For model parameter tuning, the raw data is also supplied with additional category assignment labels.

4. The image preprocessing method according to claim 1, supplies the following information:The tool provides a preformatted file containing the following data: image name, classification category, classification accuracy, and flag for manual review.The generated data is obtained by running user-provided images files through the image preprocessing chain and the model inference process.

5. The image preprocessing method according to claim 1, is packaged with a user-interface dashboard that provides the following user options:In the dashboard, the user can select the images to upload for classification, with the results generated by the model outputted to a predefined file format;In the dashboard, the user can define an accuracy threshold for processed images to be flagged for manual review, where the prediction accuracy falls below a defined threshold;In the dashboard, the user is provided statistics including: file classification summary, flagged file statistics, and total files processed.

Images