Building Training Pipelines for Computer Vision Models

This presentation will explore how state-of-the-art computer vision drives the use of autonomous systems of the future whether that’s navigating a warehouse, steering an AGV, or driving a tractor through the fields.
What You’ll learn:
* Why vision is important for all autonomous platforms, as well as the different challenges every environment creates.
* Advanced annotation & auto-labeling pipelines that dramatically speed up training-data preparation while maintaining quality.
* How to build preprocessing, augmentation, model training pipelines and deploy optimized edge models.
* Approaches related to active learning and human-in-the-loop feedback loops that allow your fleet to be continually learning and evolving.
* Best practices related to training-to-deployment orchestration using SageMaker, performance health monitoring in real time.
Whether you are designing vision systems for drones, robots or autonomous tractors, you will gain practical insights related to building data-first methodologies to accelerate your design, increase availability in various conditions, and go from prototype to your deployment fleet at scale.