Recent innovations in computer vision mean that AI can recognise images within large-scale datasets, presenting significant opportunities for innovation in just about every industry. However, this only works successfully when AI is trained with high-quality datasets, and it is difficult to organize, locate, and correct mistakes at such high scale.
Voxel51 provides a data platform for computer vision models called FiftyOne, an open-source toolkit that helps users automate the organization, navigation, and creation of computer vision models, regardless of size. The platform also includes automated data quality control tools to help users identify detrimental mistakes within models, a process that is necessary for successful computer vision.
Many startups are tackling inefficient computer vision workflows, but Voxel51 has found a gap in the market by focusing specifically on the laborious manual processes associated with organising and aggregating high-quality datasets. It has formed a collaborative partnership with Google, empowering its platform with Google's pre-annotated image datasets, and in 2023, released an AI assistant that can generate Python Code for computer vision dataset analysis. 2024 funding is being used to scale up and accelerate its AI research.
Kirsty
Company Specialist at Welcome to the Jungle
