Abstract: A case study of efficiently solving a real-world computer vision problem using a combination of labelled real-world data and synthetic data, combining the strengths of each data type. It considers best practices for combining the datasets and showcases the benefits of a platform approach using Appen's platform for real world sourcing and labelling and the Mindtech Chameleon platform to generate the synthetic data.
Bio: Dimitris is a Machine Learning engineer at Mindtech, implementing AI powered computer vision, trained on synthetic data created by their Chameleon platform. Dimitris has held a passion for AI since experiencing the fundamental breakthroughs of the early 2010’s in deep learning as an undergraduate. To pursue this interest, he undertook postgraduate studies in computer vision at Oregon State University. Aware of the limitations of real datasets he became very interested in the potential of replicating real modalities abundantly in a controlled and ethically considerate environment and joined Mindtech to substantiate the use of synthetic data by the next generation of AI.