difference between classification and detection

Question

I'm reading the following article for my master thesis: http://graphics.cs.cmu.edu/projects/discriminativePatches/discriminativePatches.pdf In section 2.1 it said: "we turn the classification step of discriminative clustering into a detection step" what is the difference between classification and detection? Initially I think it means that with "classificator" it is going to define a classifier of more classes (then exist just a classifier for all classes with input=image patch,output= class), and with "detector" a classifier of just one class (then for every class exist one different detector with input=image patch, output= yes/no). But before that line, he say "an initial clustering of data is followed by learning a descriminative classifier FOR EACH CLUSTER (class)" then, also with classifier it means "for every class (cluster) exist a classifier". then..what it would say differentiating classifier and detection? Thanks

Answer 1

Classification is a process of putting items into different bins.

Detection: Detection is a process of actually finding out about item features

Example: If i asked you to detect people coming into a room, you may have a procedure to do this and that would be detection. If I then ask you to classify them into two groups of age below 25 and above 25 you will have to do that but here most of people get confused that they have to do detection as well. They might be given information about age groups already like a list of people coming in with age and you can view that and classify them or you can use a simple detection by asking them what age they are and then classifying them.

Most of the time in image processing you will see that detection is based on looking at object and its features and detecting those features out for example detecting edges and detecting average colours. This does not classify them however you may have a classification on top of it for example classify an image into black and white and coloured picture (2 bins) or you classify image parts using edges.

Just to reiterate in other words detector is going to do its work till detection and thats it nothing else. How you use that detection is upto you. There are loads of algorithms to help in detection alone and then if you want to classify you can use neural network or simple manual classification.

When you are driving and you are looking out and detecting if there is a hazard out there. It can be a human, animal or any other vehicle etc. But If I then ask you to look out for a boy of average height then your brain would start a simple classification algorithm which would classify people into bins of male and females then reject females and apply average height classifier on males and if it finds a fit then it will trigger a Hurrah.

Answer 2

1. To my understanding, the paper applied one-vs-all SVM on multi-classification. For each cluster, the linear SVM trains to make sure the patches within the cluster truly belongs to this class (1 belong, 0 not belong). Based on the training model, the re-clustering is performed. This part is classifier.

2. The naming of detector in the paper just extended the sample from a relatively small set (discovery) to a huge set with all the possible images and patches (natural). As you can see from the sketch of Algorithm I, for each cluster, SVM training set also includes the samples from natural dataset in order to better differentiate the patches in the current cluster and those from all the other sources (other clusters and natural set). Note that although the training sample is extended, the samples for clustering won't change. The extension of training set just provides a more accurate similarity measure for further clustering.

3. The function name *detect_top* may be a little bit confused with detector, but it is just one step in the adjustment of clustering by selecting the patches from other clusters that looks more like to belong the current cluster, according to the trained classifier from previous step. From the 2nd loop, such trained classifier includes the information from the natural data set.

Classifier targets the clusters (D), while detector works for the whole data set (D+N).

Answer 3

Vision-based object detection algorithms consists typically of an image feature detector, which provides a digital description of local image content, and a pattern recognizer that classifies this local image content, e.g. as a car or background. Whereas the goal of Image classification task is to assign an input image one label from a fixed set of categories.