Course: Machine Learning for Image Data

Here is a quick recap of the material covered this week, which has focused on data and features. This week we have looked at: types of data and features feature …

In some cases it may be useful to fill in gaps in data. A constant such as zero can be used, or often the mean of the feature. However, this …

Sentinel items are specific sets of characters that are used to indicate the absence of data. This could appear as NaN (not a number) or NA (not available) depending on …

This video is a continuation of our discussion on missing data. Often, datasets are missing data for one or more features for some examples, due to human or technical error. …

This video gives an introduction to the next activity, looking at missing data. In this activity, the primary focus is on measured or derived features, rather than pixel data. In …

One of the key concepts in machine learning we addressed in the previous videos is the distinction between supervised and unsupervised learning. Here’s a summary and reminder. Supervised learning In …

Haar-like features use filters to detect regions containing areas of different image contrast – brightness and darkness in an image. It was first designed for use in face detection, and …

This is the final part of our series of videos looking at labelling image data. In this last video, we discuss the importance of quality labelling and annotation to machine …

This is part four of our series of videos looking at labelling image data. This video provides an overview of how you might label images by using filenames and folder …

This is the third in our series of videos looking at labelling image data. In this video, we look at annotating the shape of objects more accurately using polygons, and …

This video is a continuation of our series of videos looking at labelling image data. In this video, we look at labelling objects within images using points and bounding boxes.

Often, additional labelling needs to be added to images before using them to train machine learning models. This labelling provides information that the machine learning system can use to learn. …

Histogram of Oriented Gradients, or HoG is a common method of feature extraction that works on detecting edges in the image. In short, HoG constructs a histogram of edges detected …

Image data can consist of millions of pixels, which is in turn represented by millions of numbers. To reduce that amount of data into a smaller set of useful features, …

This video gives an overview of the main classes of data you might use in machine learning. This includes: numerical Boolean (true or false) frequency of appearance (e.g. words in …