PEKAT VISION is a software for industrial visual inspection and quality assurance? based on AI
PEKAT VISION utilises its own proprietary focused-learning algorithm which goes beyond deep learning. Focused-learning works like a human eye? in that it primarily focuses on details, requiring only a small set of images to understand the task.
This means it can be used to detect and classify anomalies across all kinds of materials. From wood, lacquer and leather to fabrics or metal, as final products or raw materials with uneven surfaces.
A user-friendly interface means PEKAT VISION allows for ease of use to make Deep Learning work in your application across the full Premium features:
- Unsupervised anomaly detector – Train classifiers using under positive 20 positive examples examples. Capable of inspecting stable objects as well as those with deformed and/or pattern surfaces
- Supervised surface check – A function specifically designed to find defects on completely heterogeneous surfaces, such as rust, abrasion or leakage
- Auto sensitivity – An extended part of the anomaly detection module, this determines the best sensitivity value to help find the border between a ‘good’ product and defective product
- Object detector and classifier – Find objects with non-uniform shapes regardless of orientation such as example knots in wood
- Inspection modules – Combined inspection modules to create to easily create an advanced workflow with interwoven with custom image preprocessing or scripting code
- Python Code – Utilising Python gives you high flexibility, enabling you to add advanced functions such as custom logic or preprocessing of images (e.g. with OpenCV and Numpy), or even external interfaces
- Auto-Trigger – Automatically capture images in the right moment without any necessary integration with the machine or assembly line
- Unifier – When objects need to be rotated for inspection or is different across multiple images the, unifier function normalises the position and rotation across them for easier further processing
- Preprocess module – This tool for allows for easy image editing before the next processing. It allows rotation, cropping, scaling, and background normalization
- Statistics – Calculate how successful the application evaluates the image, giving you false positive and false negative rates so you can tweak the sensitivity if needed
- Measurement – This module provides simple measurements of an objects dimensions
- Runtime statistics – View the statistics of ‘good’ and ‘bad’ images sent from the API based on specified time and dates of your choosing