New software from Teledyne FLIR expands AI at the edge
Victoria Hanscomb
Share this content
Teledyne FLIR, part of Teledyne Technologies Incorporated, has announced version 1.0 of its highly optimized image signal processing (ISP) software development kit (SDK) for embedded systems.
According to the company, its Prism ISP is designed to maximize performance in thermal and multispectral applications.
Created to run on low-power GPUs from Qualcomm and NVIDIA, Prism ISP includes image-enhancing functionality for:
- Noise reduction
- Super-resolution
- Electronic image stabilization
- Infrared (IR) and electro-optical (EO) video fusion
- Turbulence mitigation
- Locally adaptive tone mapping
Teledyne FLIR says that the Prism ISP libraries enable integrators in the defense, commercial and industrial markets to dramatically enhance imaging quality as part of AI data capture at the edge, improving performance, lowering development costs and shortening time to market.
Object detection
In addition to Prism ISP running at the edge, Teledyne FLIR says that it has also enhanced Prism AI.
This perception software enables classification, object detection and object tracking in thermal and visible light spectrums.
The company says that Prism AI now provides additional high-performance models trained on the world’s largest application-specific thermal image datasets, including ground intelligence, surveillance and reconnaissance (ISR), air-to-ground and counter-drone applications.
“Prism ISP is critical to thermal-based AI system development, offering an unprecedented performance while reducing development costs,” said Dan Walker, Vice President, Product Management, Teledyne FLIR.
“Within our end-to-end, computational imaging ecosystem featuring Prism AI, Teledyne FLIR can now support a wider array of projects and custom development that utilize low-power, embedded processors at the edge.”