Login Register



Auto-login on future visits

Forgot your password?

All videos


VISION14 talk: Machine learning for next generation vision applications

How can machine learning help us to develop next generation applications? Recording of our presentation at Industrial Vision Days conference accompanying VISION14 Trade Fair in Stuttgart, Germany.

Object detection

This video shows how to detect objects in an image using machine learning tools. First we use a classifier to detect candidate regions, and then a powerful segmentation tool to identify each object.

Detectors explained

A detector is a classifier that focuses only on one class of interest. It can be very handy especially when we have lot of samples of one specific target class and do not know much about the other classes. In this tutorial you will learn how to construct a detector in one-class scenario or when non-targets are also present.

Ground-truth by clustering

How to gather a labelled training set from images? You could paint some relevant areas manually. But this is tedious and not friendly for your wrist. This video shows how clustering may provide a better alternative.

Choose the classifier with cross-validation

How to reliably estimate the performance of a classifier? The cross-validation is a good tool for it. It can also be used to simulate a realistic classifier operation.

Tune your classifier

Not all errors are equally important. In sorting applications, some errors should be entirely avoided. In this tutorial, you will learn how to tune your classifier to find all rotten parts in French fries sorting application.

How to optimize three-class classifiers with ROC analysis

This video illustrate how to quickly optimize multi-class classifiers in imbalanced problems.

How to protect classifier from outliers

Often, we need to protect a trained classifier from accepting outliers examples appearing in production. This tutorial shows how to achieve this by addding a rejection option to a trained discriminant with sdreject command. Construction of interactive reject curve is also illustrated.