ECCV 2018 Daily - Wednesday

Daily Wednesday 12 EXPO: Mighty AI 19 Mighty AI provides labelling and dating annotation for companies in the automotive vehicle space. They take a company’s raw data, learn and understand their labelling requirements, and then deliver them a dataset with the quality of annotations guaranteed. They have a computer vision team of around 10 data scientists and are at ECCV 2018 for recruitment, to meet customers, and to understand the latest advances in the field. Colin McCarthy , the Director of Business Development at Mighty AI, tells us what makes them so special: “ We utilise our own community of 100,000 people around the world, which provides the stability to give our customers great scale and process thousands and thousands of images a day. That’s coupled with our proprietary quality management system which ensures that every annotation is of very high quality and accurate. It’s predominantly manual annotation. We utilise computer vision and machine learning to increase the speed and reduce cost on some projects. ” Colin says that the variations between customer requirements can be challenging. Different customers will have very specific ways they need their data labelled, which presents challenges both to their algorithms and to their system. They spend a lot of time learning about those requirements, setting up pilot projects, and working through different approaches to ensure that they are delivering data that is usable for their customers. They already provide high-quality annotations at a massive scale, so what is next for Mighty AI? Colin tells us that they are expanding into more automated annotations and 3D annotations. He adds that if you would like to learn more about the challenges of annotating data, they would be happy to discuss it, so do visit them at the ECCV Expo . Mighty AI have just issued a white paper with Navigant Research called ‘The Data Driving Automated Vehicles: Where It Comes From and Why Quality Matters’ which addresses some of the challenges in this industry.