Zegami Machine Learning Suite

Introducing the Zegami Machine Learning Suite, a unique set of tools to help researchers and data scientists to create, analyse and explain machine learning models, as well as, curate and annotate their training data. 

The Suite consists of two products: Analyse & Explain, which enables users to understand trends in the large datasets and evaluate the Machine Learning models that use the data, and Curate & Annotate, which is used to optimise, tag and annotate large image datasets to be used for training purposes or as useful archives. 

Analyse & Explain 

Curate Annotate

Image

Analysis

ML Model

Evaluation

Image 

Annotation

 Image Archive

Enhancement

Description

High throughput image analysis of large image data sets: extract insights, answer specific questions and qualify images for further study.

Unblack box AI models to prove validity to end users or for production use in regulated environments. 

Accelerate machine learning: use high throughput image annotation to build high quality training data sets, rapidly annotate and tag images to verify machine learning models. 

Enrich image archives: browse huge datasets at high level in exquisite detail, find rare or lost images and spot trends. Reinstate missing metadata. 

Features

  • Quality Control: investigate and analyse image datasets to spot outliers and mistagged images.
  • Collaboration: work on a data set remotely even if you have low bandwidth connection.
  • Publishing: the original image data available to anyone who has access as to the archive.
  • Machine learning: rapidly build training data sets and spot missing, incomplete or inaccurate data .
  • Explainability: automatically generate heatmaps to validate the output from machine learning models .
  • Quality Control: quickly find interesting outliers that can be missed with traditional automated methods by ‘augmenting user’s ability to see lots of information.
  • Machine learning: curate and build training data sets for training machine learning models.
  • Quality Control: Bring order to unstructured data sets by viewing and amending metadata.
  • Labelling: label large data sets with Zegami’s unique high throughput tagging method to assist in data augmentation, where little or no data exist or where data is incorrect.
  • Quality Control: bring order to unstructured data sets by viewing and amending metadata.

Benefits

  • Quickly find insights and publish findings to commercialise research faster.
  • Real time visualisation and analysis.
  • Collaborate from anywhere with our cloud based access.
  • Higher project success rate, less waste on re-running experiments.
  • Speed up quality control of large data sets by finding outliers or problems in the data.
  • Iteratively test and adjust ML models to evaluate accuracy, identify biases and the specific data causing them.
  • Exclude erroneous datapoints leading to higher quality ML models.
  • Pass regulatory requirements more easily with explainable output from models.
  • Rapidly build training data sets and spot missing, incomplete or inaccurate data leading to higher quality ML models.
  • Uncover hidden data biases.
  • Produce detailed segmentation masks to increase ML model accuracy.
  • Significantly reduce the time and resources devoted to building custom tooling, image processing systems and data pipelines. 
  • Enhance existing data sets with dimensionality reduction / unsupervised machine learning views.
  • Easily spot missing, incomplete or inaccurate metadata.
  • Integrate custom ML models to extract features from images and make data sets more queryable.

Supported Data Types

Images and metadata

Image Formats

dcm, tiff, dzi, jpg, gif, png (8 and 16 bit colour depth) and other formats upon request

Image Sources

Examples include: Infrared, Satellite, CT, MRI, X-Ray

Deployment 

Cloud

Integrations

Our flexible HTTP API allows straightforward integrations with other systems.
OAuth, 3rd Party Authentication, Single Sign On (coming soon)

Clients

Australian Plant Phenotyping FacilityHigh throughput plant phenotyping

WIMMHigh Throughput Screening (Co-localisation image analysis)

Illan Davis Lab

Understand how the nervous system works using super-resolution microscopy

Cardiovascular Clinical Research Facility (CCRF)

Retrospective studies of echocardiogram images

Leeds University/Hospital (Innovate UK grant)

Digital pathology to look at heterogeneity of staining across samples

Satellite Catapult

Natural disaster planning

Covid-19 Verification 

Prototype diagnostic tool for the detection of Covi-19 from X-Rays 

FLIR 

AI detection of objects from infra-red imaging 

Cancer Research UK 

Real-time oesophageal cancer detection in endoscopy video 

Formula One 

Key brand identification 

Satellite Applications Catapult 

Natural disaster management 

Leeds University/Hospital (Innovate UK grant) 

Digital pathology 

Leeson group 

Echocardiogram archive 

Edinburgh mouse atlas(CCRF) 

Echocardiogram repository