what is the difference between 'object detection' and 'outlier detection' in computer vision?

Question

If I'm looking at drone footage, and I'm looking for tennis courts, then I'm doing object detection.

If I decide that a tennis court is an 'outlier' as opposed to the rest of the landscape, am I now doing 'outlier detection'?

What do these 2 different names mean?

Answer 1

• Object Detection: Labelling an sample to indicate the presence of a class of object. In images, this is usually done with a bounding box or applying labels to each pixel of the image.

• Outlier Detection: Identifying examples in your data that are different to the usual distribution of yor data. Can also be referred to as Anomaly Detection.

In your example, object detection is the task you're trying to solve, with you or an algorithm working on drone footage to identify tennis courts. Outlier detection is an process that you apply to available drone recordings to identify unusual examples, compared to the "normal" images of tennis courts that you are usually detecting. These would likely contain courts that are different colours and surfaces, or look different because of damage or weather. Hope this helps!

Side Note: Identifing something as "different to the background/landscape" as you mention in your definition of "outlier" is an interesting way to approach your task. If you were trying to use machine learning to solve this task, you could try assigning a class for "Not Tennis Court" in the most simple case. Alternatively, you could try assigning more classes like "Field", "Road", "Hedge", etc. to encourage your model to distinguish between these things and possibly boost the overall performance.