Embedded Files
Working alongside FLIR’s engineering team as partners, CVEDIA designed a thermal synthetic pipeline that was able to train neural networks for a variety of use cases without the collection of more training data. CVEDIA integrated synthetic technology into FLIR’s engineering pipeline, providing FLIR with market-leading autonomous capabilities.
The pain points
FLIR, the world leader in the design and manufacturing of thermal cameras, was searching for technologies that could enable their AI roadmap. As an international company servicing both commercial and defense applications, developing innovative products was a mindset engrained in their engineering team, but the company struggled to gather enough training data for its computer vision algorithms.
As their engineering team researched data collection methods, it became clear that without synthetic data, developing many of their use cases would be impossible – there were too many data requirements, and not enough resources for extensive data collection for each one. FLIR’s CTO Pierre Boulanger researched the possibility of developing their own synthetic data pipeline, but daunted by the complexity, met with CVEDIA in 2018 and invested in the company shortly after.
"Along the lines of creating and storing thousands of images of reel - it’s so expensive, even for a successful company like FLIR, it’s too daunting. We can’t do it with real images. We have to do it with synthetic images."
Pierre Boulanger | CTO, FLIR.
Our solution
CVEDIA worked alongside FLIR as a partner over the next few years to create technology that had never been built before – software that could train thermal neural networks for a variety of different use cases, using only synthetic data.
Together, they cracked the code on domain adaptation in early 2019. The synthetic data pipeline was integrated into FLIR’s engineering pipeline, and CVEDIA technology began to service a number of applications, ranging from aerial, to animal, and maritime.
Fast forward to 2021, and FLIR is empowering the automotive community to create the next generation of ADAS and autonomous vehicle systems. Recently, FLIR released a detection algorithm to prevent road collisions with deer using CVEDIA-powered AI. CVEDIA synthetic technology was added to difficult to collect thermal deer data in order to create a robust system.
Within the defense market, FLIR is enabled by CVEDIA technology for several applications including AiTR and troop protection. Synthetic algorithms provide solutions in an industry where data collection is sparse and often not possible.
CVEDIA’s synthetic data and algorithm pipeline has been critical to ensuring FLIR maintains market leadership in the smart sensor industry, and provides strong value to its clients across commercial and defense industries.
Report abuse