Login Register

Auto-login on future visits

Forgot your password?

Newsletter February 2016

Cost sensitive optimization of multi-class ROC

How to avoid specific errors in your application?


You built a classifier that performs well in most cases. However, some of its errors are inacceptable in your application. What to do then? One possible solution is to cost-optimize your classifier. In this short tutorial, we illustrate how to use this technique to gain a fine-grained control and fulfill application-specific requirements.
Watch this video (4min) to find out more.

Acoustic detection of accidents in Alpine tunnels with perClass machine learning library

Traffic accident detections in tunnels


In a tunnel, a prompt detection of a traffic accident is crucial. It not only allows for an automatic help call, but it can also prevent more vehicles to enter and get involved. All this within seconds from the accident, way before any help could be on the location.
Such alert system, based on acoustic signal processing, has being developed by Joanneum, the top Austrian research organization. perClass was selected as a platform to develop recognition algorithms and also as the real-time deployment solution processing the live data from microphone arrays. The system, currently deployed in 10 Alpine tunnels, responds two minutes faster that the standard video-based surveillance solution. This prevents up to 70 cars (70-200 people) entering the danger zone.

Read more

Machine Learning course: 4-8 April 2016


training course on industrial machine learning

Want to learn how to build the best classifier for your problem?
In this intensive 5-day course you gain a proven methodology to solve your pattern recognition challenges. Now it includes a new lecture on Deep Learning. Lectures and hands-on exercises on industrial problems such as:

  • Image-based defect detection
  • Texture classification
  • Hyperspectral object sorting
  • Classification of database records
Join us and bring your own data to push ahead your project!

Learn more about the industrial pattern recognition course register now to the machine learning course for industry

changing color/marker style in scatter plot in machine learning problem with perClass toolbox

Set yourself markers and colors

Feature highlight

Imagine that you have few data sets with different number of classes. Wouldn't it be nice if the same class would be displayed always with the same marker and color?
In perClass you can set yourself the color and marker style for a specific class or even change the defaults anytime it is convenient. In this knowledge base article, we summarize different ways to adjust marker/color information in perClass.

Read more in the documentation