Teledyne FLIR announces the release of Prism AIMMGen
Eve Goode
Share this content
Teledyne FLIR has announced the release of Prism AIMMGen, the latest addition to the Teledyne FLIR Prism software family.
The ITAR-free AI model generation service enables the creation of artificial intelligence (AI) and machine learning (ML) models using synthetically generated data.
The new release offers cost and schedule savings for system integrators developing AI/ML products for commercial, first response and defence applications.
Separating the Prism AIMMGen from previous models
The manual labelling process for AI/ML products is a highly intensive procedure and can be heightened when it includes rare and arbitrary objects like military target classes where data collection is costly and difficult.
Prism AIMMGen doesn’t face the same struggles as previous models as it uses a synthetic data generation toolchain that is capable of generating ML-ready examples of arbitrary objects in various environments and weather in both visible light and infrared spectrums.
Generating annotated synthetic imagery in minutes that can be used to fine-tune AI models significantly reduces engineering efforts and development timelines.
“Leverage smart, synthetic data to train AI models”
Dan Walker, Vice President, Teledyne FLIR commented: “With the pace and sophistication in which systems are developing, it is critical to leverage smart, synthetic data to train AI models so they are best-in-class.”
He added: “AIMMGen speeds developers’ time to market while ensuring that their respective AI products are ready for whatever scenario they might encounter in practically any environmental condition.”
Cost effectiveness
The automated AI optimisation system produces high-quality models without costly data collection.
Its ML operations (MLOps) infrastructure ensures data safety and leverages an extensive synthetic data lake and expertise that Teledyne claims has taken decades.
This development is cost-effective and enables new task requirements for the software including target classes or fine-tuning.
Prism AIMMGen features
- Unlimited synthetic images – It gives users access to a large ITAR-free data lake for multiple object detection use cases.
- AI training datasets – This features gives users the ability to produce millions of automatically labelled synthetic image sets to meet specific application requirements ensuring efficient, focused learning, including an interactive search tool.
- Model training, validation and optimisation – Its short-lead generation of AI/ML solution is at a fraction of the traditional engineering effort on private MLOps cloud infrastructure.