August 2020 Summaries

Voxel51 has launched FiftyOne, an open-source tool designed to enhance the quality and diversity of image datasets used in machine learning, addressing the challenge of performance-limiting datasets. By providing an interactive dashboard and Python library, FiftyOne allows data scientists to visualize, explore, and analyze datasets efficiently, reducing the time and effort required for continuous data quality management. This tool supports multiple image dataset types and integrates with popular machine learning frameworks like TensorFlow and PyTorch, enabling users to curate superior datasets by removing duplicates, correcting labeling errors, and identifying unique samples. Founded by Dr. Jason Corso and Dr. Brian Moore in 2016 and based in Ann Arbor, Michigan, Voxel51 aims to transform dataset curation into a science, thereby improving the predictive performance of machine learning models.