image processing - Computer Vision - Is it necessary to have multi classifiers with certain viewpoint for object detection? -

let want train hog descriptor + linear svm car detection. necessary me make, let 3 classifiers, back-view, front-view , side-view of car or can train single classier viewpoints of car?

it's not necessary recommended. can make single classifier handles multiple cases won't perform overall. issue here isn't variability of descriptor responses between different views, difference in aspect ratios between rear/front-facing , side-facing detectors. sliding window use extract hog either capture negative (sideview-sized on rear/front) or not enough positive data (rear/front-sized on sideview).

bottom line: depends on accuracy/processing rate requirements. experience front/rear similar enough, if high accuracy desired, you'll need separate detector each. need separate side detector , might need dedicated 'quarter view' detectors - front right, front left, rear right, rear left.

p.s. i'm omitting fact need handle multiple scales.