HOG feature extraction for 112x92

person = 5; [hogFeature, visualization]= ... extractHOGFeatures(read(training(person),1)); figure This is the code for face detection. The code extracts the first image of a person. the image size is 112x92. In the workspace the hog feature shows 1x4680, for different dimension of image the hog feature value changes. I am unable to understand the changes of HOG feature. Can anyone help me please

Answer 1

[Jan]

Did you read the corresponding docs already? See: https://www.mathworks.com/help/vision/ref/extracthogfeatures.html#outputarg_features . Which detail is not clear yet? It would not be efficient to rephrase the complete help text.

[Tousif Ahmed]

BlocksPerImage = floor((size(I)./CellSize – BlockSize)./(BlockSize – BlockOverlap) + 1)

Image size is 112x92, Cell size is 8x8; block size is 4; blockoverlap is 2. I know the answer for blockperimage is 130 but cannot get that by using the formula.

[georges matar]

because you are using blocksize as a number 4,,, instead, replace blockSize by the array [2 2]