Algolux specializes in smart optimization for vision systems to significantly improve development time, cost, and accuracy. The company recently announced CRISP™-ML, the first machine learning platform to automate the complex optimization of vision systems. CRISP-ML applies an innovative machine learning solver approach to intelligently find the best case set of tuning parameters against targets set by the development team. Below is our interview with Allan Benchetrit, President and CEO of Algolux:
Q: Allan, tell us something more about the company and your core competence?
A: Algolux specializes in smart optimization for vision systems, leveraging the deep expertise in image processing, computer vision, and machine learning of our PhD researchers and a robust patent portfolio into our unique platforms. We are based in Montreal, with additional staff in Silicon Valley, enabling us to be at the forefront of research and engineering in these technologies.
Q: You’ve recently announced CRISP™-ML, the first machine learning platform to automate the complex optimization of vision systems; could you tell us something more?
A: A significant portion of the design process for a camera-based system is the optimization, or tuning, of those systems for image quality, vision accuracy, or a combination of both based on the type of system. For example, an ADAS or autonomous car vision system, a high quality mobile camera, or an industrial inspection system have different requirements. Typically, a team will evaluate hundreds of tuning parameters against a large set of standard test and application specific images, which could take many calendar months for the team. CRISP-ML applies an innovative machine learning solver approach to intelligently find the best case set of parameters against targets set by the development team, a process that is mostly done manually today as there are no commercial solutions available. The platform can be deployed through a traditional on-premises use model or as a service in the cloud. You can see more at www.algolux.com.
Recommended: IRYStec’s Perceptual Display Platform Embedded Software Technology Replicates How The Human Eye Sees
Q: What are main benefits of your platform?
A: CRISP-ML significantly shrinks imaging and vision system development time and cost while improving accuracy. Today, this optimization process is very intensive and manual. In our discussions with these teams, it is not unusual for them to spend 4-6 months getting this right, and some have told us they spend up to a year optimizing just image quality. Shrinking the time down to days or weeks helps them to get to market much faster. The efficiency of this approach also allows teams to evaluate more combinations of optics, sensors, and algorithms to reduce system BOM cost or power. This provides direct business benefit for both the vision system providers and the integrators.
Q: What markets can take advantage of what Algolux provides?
A: We’re at the forefront of explosive growth in the computer vision market that is driven by incredible innovation across many markets and applications. Market research firm Tractica estimates the computer vision market will grow to nearly $50B by 2022, a CAGR of 33%. ADAS and autonomous driving, drones and AR/VR systems, robotics and machine vision, and security and surveillance make up the majority of that. Each application will require specific tuning to achieve the required image quality and vision accuracy for a viable product, but the expert resources are just not available to do this manually today. Our ability to automate the optimization process and abstract the complexity with CRISP-ML creates a significant opportunity for both customers and Algolux.
Recommended: Plotagraph Pro – Animate Any, Single, Still Image – From Any Era
Q: What are your plans for next six months?
A: We are currently involved in pilot engagements with both CRISP-ML and CRISP-EC, our auto-tuning image signal processor (ISP) product, and will expand the engagement program with customers in key markets. We’ll continue to focus on ease of use to better scale our ability to grow and support our customer base. In order to execute on these plans, and given the strategic nature of our technology and expertise, we are engaged in funding discussions with various industry partners.
Recommended: Wings – A Decentralized Platform To Create, Join And Manage DAOs
Q: More generally, how do you see the machine learning landscape developing?
A: Similar to computer vision, machine learning and especially deep learning approaches have moved from research into commercial applications at a very fast pace. The results of these approaches have been impressive and continue to improve as processing power increases and allows the training and deployment of ever deeper neural nets. In addition, frameworks such as TensorFlow and Caffe are making it easier to experiment with deep learning thereby opening access to many more people. Although we’re still very early in the evolution and application of these techniques, we are very excited to see how the technology evolves to enable new capabilities. We’ve already seen great results and that is why Algolux continues to invest in this area.Activate Social Media: