Semantic Segmentation - Standardizing Labeling of Ground Truth Datasets
Traffic laws are arguably one of the most wide-spread, common applications of standardized rules across complex, real world scenarios. In order to participate safely and anticipate the behavior of others, individuals generally decide to surrender their own perception of what is right and willingly abide by greater rules created by society. The presenter will argue that semantic segmentation of ground truth data can and should be approached in a similar way. Through standardization of taxonomies and delivery formats, high in-demand training data sets can be created in more intuitive ways, while also enabling faster turnaround times and increased focus on precision.
Emanuel Ott has been working as a Solutions Architect at iMerit for over 3 years leading iMerit’s Computer Vision Dataset Services. He and his US/India based teams have been providing the world’s leading augmented reality and self-driving car focused companies with training data to power their Machine Learning efforts to “see the world”. Emanuel has extensive experience aiding clients in drafting and refinement of guidelines, improving data pipelines, and identifying way to leverage and improve human-in-the-loop workflows to be most effective.