“Squinting Vision Pipelines: Detecting and Correcting Errors in Vision Models at Runtime,” a Presentation from Squint AI

For the full video of this presentation, please visit: https://www.edge-ai-vision.com/2025/06/squinting-vision-pipelines-detecting-and-correcting-errors-in-vision-models-at-runtime-a-presentation-from-squint-ai/ Ken Wenger, Chief Technology Officer at Squint AI, presents the “Squinting Vision Pipelines: Detecting and Correcting Errors in Vision Models at Runtime” tutorial at the May 2025 Embedded Vision Summit. As humans, when we look at a scene our first impressions are sometimes wrong; we need to take a second look, to squint and reassess. Squinting enables us to focus our attention on the subject we are investigating and often clarifies and corrects our initial assumptions. Computer vision algorithms, including those enabled through artificial intelligence and machine learning pipelines, can also be made to squint. As the role of AI/ML algorithms in automation matures, the cost of mistakes will increase dramatically. Consider use cases in the healthcare industry, cancer diagnosis and research and autonomous systems. In this presentation, Wenger shows how the Squint Insights Studio platform uses explainable AI to add context and reasoning to vision model decisions, enabling the user to build vision pipelines that continuously assess and revise vision model predictions in production environments.