Title: Data-driven Advancement for Autonomous Driving
Summary: Autonomous cars have an massive likely to preserve life and minimize greenhouse gas emissions, as properly as generating transportation additional adaptable and relaxed. Machine learning is a important tool that enables autonomous cars to perceive their ecosystem and learn from a continually rising system of driving information. In this speak I existing how we create data-driven notion devices at Motional. Aside from presenting our percep- tion algorithms (PointPillars, PointPainting) and general public benchmark datasets (nuScenes, nuImages), I go over how to create authentic-globe machine learning solutions. A distinct concentrate will be on the areas that academia are unable to address for us: deciding upon the correct information utilizing Lively Studying, defining what to annotate and scaling the pipeline up to formerly unseen quantities of details.
Come across out extra at https://idsc.ethz.ch/investigation-frazzoli/autonomy-talks.html
