On February 6, 2024 at 9 am PT (noon ET), Julie Buquet, AI Scientist for Imaging, and Ludimila Centeno, Associate VP of Technology Offer and Support, both of Immervision, will present the free hour webinar “Optimizing Camera Design for Machine Perception Via End-to-end Camera Simulation,” organized by the Edge AI and Vision Alliance. Here’s the description, from the event registration page:
Camera design is an iterative process that minimizes complex cost functions while optimizing the optics, sensor and image signal processor (ISP) for multiple interconnected parameters. New types of lenses (such as metalenses and free-form lenses), along with new types of image sensors and improved image processing techniques, offer many ways to optimize a camera for an application.
But what criteria are we optimizing for? Camera designers today still typically focus on key performance indicators (KPIs) derived from human perception (e.g., removing aberrations and increasing sharpness). But in a growing range of applications, images are used for computer vision rather than human viewing. Camera design KPIs need to be redefined for best results in machine perception applications.
In this webinar, Buquet and Centeno will explain how to optimize camera performance for machine perception applications through simulation and end-to-end design. Computer vision requirements are often initially based on algorithm performance targets. Upfront translation of these objectives into camera KPIs is necessary to support the design process. Then, by simulating the entire camera—optics, sensor, ISP and other factors—it is possible to estimate the impact of each component on machine perception performance even prior to building a physical prototype. This approach also allows for automated optimization of optical parameters.
Buquet and Centeno will step through the imaging pipeline, describing various stages including commonly undocumented ones such as image quality tuning done by the ISP. They will highlight how Immervision has built complex cameras more efficiently while meeting machine perception performance requirements at reasonable cost. Buquet and Centeno will use a wide-angle in-cabin camera case study to show how these techniques play out in applications. A question-and-answer session will follow the presentation.
To register for this free webinar, please see the event page. For more information, please email [email protected].
