An online marketplace that lets developers buy data directly from members of the public in a clear and responsible way has been launched by Microsoft.
Trove, which recently launched in the UK, allows researchers to publicise the artificial intelligence projects they are working on and ask people to contribute photos to help improve their machine learning algorithms.
Developers and researchers rely on large amounts of high-quality and diverse photos to train computer vision models, which are used in a range of tools and services – from self-driving cars to cameras that help people with blindness understand what’s around them.
However, these images are often crowd-sourced by people who have little to no insight into the purpose of the project and the exact types of images that are needed. This can have a serious impact on how the tool or service performs.
Trove solves this by ensuring developers can ask members of the public for the exact type of imagery they need for their project.
Everyone who has an image approved by a developer in Trove is paid for that photo. They can also read a clear and easy-to-understand description of the purpose of the project, directly communicate with developers to ask questions and decide how their photos are used.
Ashly Yeo, Senior Program Manager at Microsoft, said: “Most of the projects on Trove are for business solutions. A supermarket could benefit from a camera pointed at its shelves that can ‘see’ when a product has sold out and alert staff so they can refill that space. For the camera to recognise the products, it needs a machine learning algorithm that has been created using a lot of different images of those products.
“Developers need very specific imagery and this is what Trove can help provide, while building trusted connections between the people who depend on data and the people who have it.”
Microsoft deletes information from uploaded photos that could potentially identify individuals, such as where the image was taken. Security and privacy around data is also a key consideration, with only the developer who created the project and their team able to access the images.
Developers who want to use Trove have to complete an onboarding process that asks them about the data that will be collected, how it will be used, examples of data they are looking for, how many photos they want to collect and how much will they pay data providers. They are also given a guide on the rules around collecting data from the public.
To learn more about Microsoft’s AI principles, click here.