Over the last 5 years the Sterblue team has built an AI-powered cloud platform to support the largest energy companies in the world (EDP, E.ON, American Electric Power, Enedis, ESB, etc.) in managing their critical assets: wind turbines, transmission & distribution grids, cooling towers. Machine learning was a huge part of the product.
Millions of images have been processed on the platform. AI was used extensively to detect discrepancies (corrosion, cracks, vegetation, etc.), to locate assets in space, to generate automatic flight paths for drones, and many other applications.
5 years ago, showing bounding boxes on an image on a Linkedin or blog post was the big marketing trend. Sterblue chose to do it the hard way, by scaling up with AI.
“We labeled millions of images and worked with a number of different labelling companies. We also dealt with tens of different taxonomies for the same physical problem, we faced labelling quality challenges, we built AI training pipelines using tens of open source frameworks and ecosystem partners, we integrated real users in the loop. We really felt some pain here to easily unleash AI’s full potential for the world” said Geoffrey Vancassel, CEO and co-Founder at LabelFlow
“At LabelFlow, we want to support every AI-applied company in the world to build the next big thing. We’re starting small with our labelling tool, with a huge focus on user experience and ecosystem openness. You can get the tool tested in 2 min with no login required … and start building your data set. LabelFlow becomes your single source of truth.” Added Nicolas Draber, LabeFlow's CPO and co-Founder
What’s new as compared to other players on the market:
- Smooth user experience: Give it a try, no login required! Priority has been given to user experience and to an easy onboarding process.
- Blazing fast: LabelFlow is designed for optimum productivity. Keyboard shortcuts, interface layout and everything else is designed with users, for users (with collaboration on the way!)
- Data ownership: LabelFlow does not try to own your data or algorithms, but integrates with them seamlessly. No duplicate source of truth and complicated scripts to synchronize your data between various tools. Your database is the only source of truth for LabelFlow.
- Open community and standard: Our mission is to generalize standard good practices around visual data management. Dataset curation should not require any "secret sauce". LabelFlow is building a community around an open labeling tool to set the standard.
LabelFlow builds the open platform for image labelling to support building the next big thing with AI. LabelFlow allows data scientists, AI startups & corporations to properly create and manage their datasets to bring computer-vision based AI to scale. LabelFlow technology was used over the last 5 years to label 1.5m images helping the largest energy companies in the world to monitor their critical assets.
+336 68 669 484