V7

About us

V7 is a platform for the management and labelling of machine learning training data, and the training of machine learning models for the automated completion of visual tasks. With over 1 billion labelled  training images under management, V7 allows over 300 AI companies and research groups with the creation of AI systems that are accurate and medically compliant.

Headquartered in London, V7 was founded by Simon Edwardsson and Alberto Rizzoli in 2018 to enable artificial intelligence projects to control their “ML-Ops” in a platform with good enough UX to enable clinicians to perform image labelling, and technically versatile enough to allow data scientists to run neural networks within it to assist the annotation process.

V7 was named by Forbes as one of the 20 Machine Learning companies to watch in 2021. Today it is responsible for tens of thousands of accurate medical diagnoses every month.

Learn more about us at  https://v7labs.com.

Main role in the project

V7’s platform will be used to organize, label, and visualize the iTobos training data. These images will become available in a searchable, user-friendly dataset interface. Data scientists will be able to tag, sort, and segment the data to analyse sub-datasets, detect potential bias, and test model accuracies on specific validation sets that accurately represent a population’s appearance.

New annotation capabilities within the platform will be used by labellers to detect and segment moles. V7’s auto-annotation system will be used to semi-automatically segment moles and other skin lesions thereby increasing the amount of training data that the iTobos system can leverage for each hour spent labelling. The platform will keep track of who labelled, and which clinician reviewed an annotation, maintaining a secure history of each component of the project’s training data down to a single bounding box or polygon label.

Finally, V7’s deep learning researchers will collaborate in the creation of mole detection and segmentation models that can be used both in model-assisted labelling and in production.