The key to a AI project is data. Microsoft’s latest GarageTrove app hopes to bridge the gap between photographers and data scientists, who can use the images they take.
The Trove app is currently available on Android and iOS platforms, where users can submit photos to contribute to AI projects, which developers can use to train learning models.
Developers and researchers rely on a large number of high-quality and diverse photos to train their computer vision models, often by crowdfunding of photos they collect, asking workers to help submit photos.
But there is a problem where the photo submitter knows almost nothing about the project’s larger context or purpose, and when the photo doesn’t meet the project’s needs, their work may be rejected without explanation.
To do this, Trove addresses this need, connecting contributors and AI developers, and putting communication, background, and feedback first.
The main features of Trove include
Contributors can submit existing photos or actively take new photos and have a chance to win the lottery.
View AI items that require photos
Understand the purpose of the project and the expected parameters for submitting the photo.
Provide photos for your favorite projects
Direct communication with developers, asking questions or receiving feedback
Review project owners and contribute to building a people-centred community
Earn a raffle by contributing and recommending other contributors.
Developers can request photos of AI projects by listing them in the APP.
Get photos for your AI project
Inspire enthusiasm for your project by explaining your modeling purposes and providing participation in the raffle.
Provide feedback directly to contributors to help them re-adjust their submitted photos to match your project parameters.
:: Review and thank contributors
Microsoft says Trove is rooted in fair trading, has no hidden fees, and offers licenses that will protect your image ownership. The result is a responsible data source that provides developers and contributors with more choice, control and security.