With the ability to quickly automate complex and subjective decision-making, deep learning solutions for production lines can help identify defects, ensure the quality of goods, increase productivity and create process efficiencies. Despite its utility, implementing deep learning in an industrial environment can often be intimidating due to the number of tools, large datasets, specialized skills, and associated costs.
To empower those with no AI expertise and to reduce the complexity and cost of implementing deep learning inspection on the production line, FLIR Systems has collaborated with Neurala to develop a fully supported software and hardware solution. This easy-to-deploy solution saves time as it helps prototype, train and validate models without needing to setup a traditional AI development environment.
Using Neurala Brain Builder module of the Visual Inspection Automation (VIA) software, customers can annotate, label, train and validate neural networks in minutes. These models can be directly uploaded to FLIR Firefly DL machine vision cameras using the free FLIR Spinnaker SDK. The solution delivers a seamless workflow that replaces reliance on additional third-party tools for annotating, labeling, training, and validating neural networks; achieving similar results at one-third of the cost.
Coming at a time when manufacturers are increasingly adopting AI and automation to address challenges like workforce availability and supply chain disruptions, the solution can be configured for a variety of automated classification-based applications at inspection points. Use cases range from printed circuit board (PCB) inspection at a semiconductor fabrication plant to fruit sorting and grading at an agricultural processing plant.
Machine vision builders, system designers, and those getting into deep learning alike now have an easy-to-use, time-saving, and cost-effective method for automating classification-based inspection on production lines.
Neurala’s VIA software for FLIR Firefly DL is available in the US today and will be available globally mid-year 2021.