Drives and Controls Magazine
Deep learning vision camera runs on-board neural networks
Published:  21 October, 2019

Flir Systems has developed what it describes as “the industry’s first deep-learning, inference-enabled machine vision camera with Flir Neuro technology”. The Firefly DL camera combines machine vision with deep learning to tackle complex and subjective applications such as recognising faces or determining the quality of solar panels.

The on-board Neuro technology allows users to deploy trained neural networks directly on the camera, eliminating the need for a host system when carrying out tasks such as classification, object detection and localisation. These networks support on-camera decision-making and make “inference on the edge” possible.

The open-platform Neuro technology supports popular frameworks such as TensorFlow and Caffe.

A trained neural network can be deployed on Flir’s deep-learning camera

Flir says that the camera – which costs £231 (ex VAT) in the UK – will allow original equipment manufacturers, engineers and manufacturers to solve challenging automation problems quickly. The small, low weight and low power consumption device can be embedded into mobile, desktop, and handheld systems.