I am reading the following article for my master's thesis: http://graphics.cs.cmu.edu/projects/discriminativePatches/discriminativePatches.pdf In section 2.1, he said: “we are turning the discriminatory clustering classification stage into a detection stage”, what is the difference between classification and detection? At first, I think this means that with the “classifier” it will determine the classifier of more classes (then there is only a classifier for all classes with input = image patch, output = class), and with the “detector” - the classifier of only one class ( then for each class there is one other detector with a patch input = image, output = yes / no). But before this line, he says: “The initial clustering of data is accompanied by the study of the descriptive classifier FOR EACH CLUSTER (class)”, then also with the classifier it means “for each class (cluster) there is a classifier”. then ... what would he saydifferentiating classifier and detection? Thanks
As far as I understand, the document uses SVM with one vs-all for multi-level classification. For each cluster, linear SVMs are trained to make sure that corrections within the cluster really belong to this class (1 belongs, 0 does not belong). Based on the learning model, re-clustering is performed. This part is a classifier .
() (). I, SVM , ( ), , , . .
* detect_top * , , , , . 2- .
(D), (D + N).
- .
: - .
. , , , . 25 25 , , , . , , , , , , , .
, , , . , , , - (2 ) .
, , , . , , . , , , , .
, , . , .. , , , , , .
, , , . . , .