Training-ready data packs to accelerate AI-based computer vision systems
Launched in 2019, Mindtech’s behavioural simulator platform, called Chameleon, creates synthetic data to train AI-based vision systems faster and more accurately than other solutions. Mindtech customers, across a range of sectors including retail, home automation and security, use Chameleon to rapidly create the training data they need for a range of tasks including control and fault detection, as well as safety and product recognition.
Building upon the success of its Chameleon platform, Mindtech is now offering customers synthetic training-ready data packs, usable both as a standalone product or as part of the Chameleon platform. Framework agnostic, these data packs provide high fidelity images and image sequences, alongside advanced and accurate annotations such as instance segmentation and skeleton key points.
Able to be used immediately, the training data pack ensures GDPR compliance through its use of synthetically generated photo-realistic humans and includes annotations and data that can speed up the training of AI-based computer vision systems significantly. For instance, in the time it takes to gather 10,000 real world annotated images, Mindtech’s data packs can produce 1M compliant images.
Additionally, the training data packs have been carefully curated in order to address the most common customer needs including the inclusion of appropriate camera angles, camera simulations and lighting. Enhanced privacy features mean that any photo-realistic humans are synthetically generated, so that users can adopt the platform with ease.
Steve Harris, CEO at Mindtech Global, commented: “When it comes to building AI-based vision systems, real-world data is scarce, time consuming to annotate, and brings data provenance challenges. Our Chameleon platform empowers users to create high-quality, precisely annotated and privacy compliant synthetic data quickly and easily. And, with our new data packs, customers can purchase readily-available synthetic data which fits their needs.”
2022-10-11
