We have experience building computer vision solutions in several industries and deploying for desktop and mobile app based environments.
In particular we have worked on computer vision applications for biometric security and am listed as an author on one patent regarding security of biometric templates.
In recent years Convolutional Neural Networks (CNNs) have eclipsed traditional computer vision techniques for a variety of tasks including image classification, face recognition, object detection and localisation, and more. The algorithms are efficient but computationally intensive and many CNNs require bespoke hardware such as Graphical Processing Units (GPUs).
CNN solutions also require collecting, sanitising and augmenting large training datasets. For example you may need 10 million faces or more to train a face classifier.
We can build a custom CNN solution for your business and can assist with obtaining, gathering and sanitising training data, choices of network architecture, performance considerations on different devices, and any extra GPU hardware that is needed.
Computer vision technologies that we have worked with include
- Cascade classifiers
- Nvidia GPUs on site
- Google/AWS/cloud hosted GPU instances