Algolux, Inc. today announced CRISP-ML, the first machine learning platform to automate the complex optimization of vision systems. The platform’s advanced technology and architecture provide a unique advantage to both experienced and novice computer vision teams in addressing the growing cost, expertise, and time to market challenges they face in optimizing their systems.
The number of companies integrating cameras and intelligent vision technology into their products is increasing rapidly, and the image quality and accuracy of those systems is core to their value. Improving the ability to avoid a collision, identify a person, or recognize a manufacturing defect is a critical way to differentiate from their competition and build trust with their customers. The payoff will be significant: According to market research firm Tractica, the computer vision hardware and software market is expected to grow to $48.6 billion by the year 2022 from $6.6 billion last year, a CAGR of 33 percent.
Optimizing these systems across combinations of optics, sensors, processors, and algorithms requires intensive manual effort. This painstaking work must be done for each product and variant, severely limiting the number of new configurations that can be evaluated. In addition, top-tier providers struggle with growing costs and schedule overruns while most smaller providers can’t attract the required deep expertise, often outsourcing and giving up critical control. These challenges continue to grow as the systems and underlying algorithms become more complex, fundamentally driving the need for a new approach.
“Recent advances in vision technologies are transforming many markets and enabling impressive growth opportunities, but developing an optimized vision system with the necessary accuracy and image quality can be very complex and resource intensive,” said Jeff Bier, founder of the Embedded Vision Alliance. “I applaud Algolux’s innovative work towards reducing cost and time to market, as well as helping address the expertise gap many companies face when developing these systems.”
A New Paradigm for Optimizing Vision Systems
In collaboration with leading universities and leveraging its own deep expertise in imaging, computer vision, and machine learning, Algolux has architected an optimization platform based on its innovative machine learning solver that holistically tunes the imaging and computer vision algorithms based on standard image test charts, tagged training images, and Key Performance Indicator (KPI) targets to achieve the required image quality, vision accuracy, power, and performance.
As the solver iterates, it generates reports of its progress toward achieving these KPIs, and outputs image files with data overlays to compare against the known good input images. Once KPIs are achieved, CRISP-ML exports the optimal parameters for each system configuration of optics, sensors, and algorithms tested.
The design team guides the optimization process through the management console to integrate and configure CRISP-ML to the target vision system, manage and launch multiple KPI scenarios, warehouse and analyze reports and images, and enable local or cloud-based runtime to best leverage engineering and processing resources across geographies.
- Save many hundred thousand dollars per vision system per new product
- Accelerate testing and optimization by an order of magnitude to reduce time to market
- Improve leverage of existing engineering expertise and resources
- Automate testing of many more component combinations than possible before
- Achieve optimal vision system cost, power, and performance
“The explosive growth of camera and vision systems being integrated into products is enabling many new markets and invigorating traditional ones. A tremendous opportunity awaits successful providers but achieving the optimizations required to maximize commercial potential is out of reach for many companies,” said Allan Benchetrit, President and CEO, Algolux. “CRISP-ML combines our image processing and computer vision expertise with our advanced machine learning technology to automate and accelerate the holistic tuning of these vision systems, achieving better quality of results at lower cost and much faster than traditional methods.”
CRISP-ML is currently in limited access, with broader availability in Q1 2017.