Machine vision has been a staple for industrial manufacturing. The capability is deployed in various scenarios, including quality inspection, object, and defect identification. In recent years, machine vision has also been crucial in the rise of the autonomous mobile robot, as machine vision plays a key role in visual based simultaneous localization and mapping.
With the rise of deep learning, more and more machine vision models are built on deep learning techniques, predominantly on convolutional neural networks. By building on convolutional neural networks, the accuracy and capabilities of deep learning based machine vision will improve as more data are gathered and used to train the model. One of the main verticals that are looking to adopt deep learning based machine vision is the manufacturing industry. As the manufacturing industry starts to embark on the journey of digital transformation, the adoption of deep learning based machine vision is expected to grow. Instrumental, Landing.ai and micropsi industries are among the main startups that are offering this technology.
At the moment, the implementation of deep learning based machine vision will rely heavily on edge device and on-premise servers. As the technology is still in its early stage, implementers need to be aware of several requirements prior to deployment. The training and testing of deep learning based machine vision models and algorithms require high definition data and high computational power, which is currently lacking in the manufacturing environment. It is also critical for the models to be interoperable with existing infrastructure, such as industrial-grade camera from Cognex, Basler and Keyence, and industrial cloud platform from industrial players like GE, ABB, SAP, PTC and Siemens. Redundancy, data privacy and sovereignty requirements should not be overlooked as well.
Table of Contents: