As AI expands beyond content generation to perceiving and understanding the physical world, intelligent sensing is becoming increasingly important across industrial applications. At CIOE 2026, Raytron, a global leader in infrared thermal imaging, unveiled its Infrared Vision-Language Model, extending infrared imaging beyond image acquisition toward intelligent understanding. Raytron also introduced Falcon 500, its third-generation AI image processing chip, and showcased the latest advancements in its 8 μm and 12 μm infrared detector portfolios.
Why Does Infrared Imaging Need a Dedicated Vision-Language Model?
Unlike visible-light imaging, which primarily captures reflected light, infrared thermal imaging detects radiation emitted by objects, revealing unique information such as temperature, heat distribution, and thermal signatures. Most general-purpose vision models are trained primarily on large-scale visible-light data and are not specifically designed to interpret the unique characteristics of infrared imagery. Raytron’s Infrared Vision-Language Model is designed to address this challenge through:
5M+ Proprietary Infrared Data: More than five million high-quality multimodal infrared data samples cover specialized scenarios including industrial temperature measurement, night vision, and gas detection.
Dedicated Infrared Visual Encoder: Designed specifically for infrared imaging, the dedicated encoder enhances the model’s ability to capture and interpret thermal distributions and infrared signal features.
Industry-Specific Training: The model is trained for application-specific scenarios, such as gas detection, electrical inspection and perimeter monitoring. Training data covers gases such as methane, sulfur dioxide, ethanol, R-134a, R-152a, and sulfur hexafluoride, while accounting for gas plume patterns, dispersion behavior, and background interference to improve the identification of potential leakage areas.
Edge AI Deployment: Enables local deployment and hardware acceleration of AI models on edge chips, allowing devices to perform real-time intelligent sensing without relying on cloud connectivity. This reduces latency, enhances data security, and helps optimize overall hardware costs.
Advancing Infrared Sensing with Falcon 500 and 8 μm Infrared Detectors
The launch was accompanied by further advances in infrared image processing and detectors. The company unveiled Falcon 500, its third-generation infrared AI image processing chip, bringing enhanced image quality, intelligent sensing, multimodal fusion, and precision temperature measurement.
At the exhibition, Raytron also showcased its expanded 8 μm lineup, spanning resolutions from 640×512 to 1280×1024, alongside upgraded 12 μm products offering NETD as low as 15 mK.
About Raytron
Raytron is a leading global supplier of core components and subsystems for multimodal sensing, with expertise in infrared, microwave, laser sensing, and AI-powered perception. Its full-stack capabilities span core chips, detectors, camera cores, complete devices, and AI algorithms, supporting system-level solutions for professional applications.
