Matrox Imaging has announced a software update for its vision software, Matrox Imaging Library (MIL) X, with two updates. The software development kit (SDK) includes a collection of tools for developing ‘machine vision’ applications.
The MIL X service pack includes a range of new features and functionality including the training of deep neural networks for image-oriented classification, coarse segmentation using image-oriented classification based on deep learning with a revamped plus augmented 3D display, processing and analysis as well as support for HDR imaging.
An additional companion update introduces enhancements to the MIL CoPilot interactive environment such as support for training a deep neural network.
The MIL X service pack also includes expanded capabilities of its classification tools using deep learning technology such as convolutional neural networks (CNN’s) to analyse images of textured, varying and ‘acceptably deformed’ goods.
MIL X includes infrastructure to build a training dataset such as labelling of images and augmenting a dataset with synthesized images with monitoring and analysing the training process.
Multiple types of training are supported, such as transfer learning and fine-tuning.
Users can also train a CNN individually or liaise with Matrox Imaging’s team to perform the training on a user’s behalf via Matrox Professional Services.
MIL X’s image-oriented classification makes use of deep learning technology in two ways, with a global approach that assigns images to classes and a coarse segmentation approach, which is included in the new service pack.
The coarse segmentation approach maps image ‘neighbourhoods’ according to categories, identifying and roughly locating the presence of features and defects.
The registration toolset has also been expanded, adding a feature for HDR imaging alongside the MIL CoPilot interactive environment for experimenting, prototyping and generating code. A companion update to the service pack adds training and interference support for image-oriented classification using deep learning.