FEATURE Machine vision
Properly training your machine- vision system
By Miron Shtiglitz, Director of Product Management at QualiSense M
achine vision for parts inspection is a technology beset with misinformation. Just
five years ago the only way to teach a machine-vision system required every possible fault, mark and annotatation of errors to train the system with. This is labourious process, done during the commissioning stage by the vision system integrator and kept up to date by the quality assurance manager. In 2017, deep learning, autonomous machine vision and artificial intelligence became the industry’s buzzwords, with some vendors offering supposedly much simpler ways of training a machine vision system. But, unsupervised deep learning such as this, where the model learns only from ‘OK’ samples, struggles to deliver the desired results. In most cases, 20, 50 or even 100 OK samples can’t really represent the entire diversity of a product, unless the product is very simple or the inspection task trivial. Today’s supervised deep-learning systems require hundreds or thousands of annotated images to deliver an adequate solution. A system will not
18 March 2024 | Automation
learn from 20 or 50 OK samples, but most likely show false positives, and even missed defects.
The alternative
QualiSense’s Augmented AI Platform only requires access to unlabelled production data and images, which are both available in abundance and require no manual tagging. It uses its AI engines to process the data and to automatically sort it. QualiSense uses smart user feedback to sharpen its tools and constantly improve the result. When data is initially sorted it goes into an auto- labelling engine annotates the defect, followed by smart feedback to tweak the system, if needed. Furthermore, QualiseSense can adapt the model to any process or environmental change in real time, meaning that you don’t have to worry about your system not keeping up to date.
Along with other advantages, such as minimal data handling, simple integration and quick installation and commissioning, QualiSense’s software-
only solution runs on almost all machine-vision systems, with the help of the QualiSense API. This includes the ability to retrofit a failed installation sitting idle on a production line.
Failed installations?
The global market for machine vision is estimated at around $15bn, with each traditional installation costing around $150,000. Yet, many quality assurance managers state that a significant percentage of installations don’t work. If this number of failed installations is as low as 10%, then we are looking at some 10,000 new points of vision every year that are not functional. That tells us that over the lifetime of the machine vision market there are, conservatively, hundreds of thousands of hardware systems spread around the world, waiting to be brought to life. Imagine the impact on the global economy, and on the output of your own plant, if that were to happen. So it makes good business sense to find the right system and its training model than blindly trust news about the number of good images your system needs.
automationmagazine.co.uk
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42
