java - The implementation of HOG Descriptor -
i trying implement hog descriptor using java without using hog descriptor implemented in opencv, have following questions:
are hog detector , hog descriptor same?
after extract interest points or key points image extract hog descriptor? form whole image or interest points / key points detected before image.
when compute gradient divide image blocks before or after compute image gradient?
regards in advance
well,
- i haven't used hog judging other descriptors, not same. actual feature descriptor while detector can guess used detect (to locate) feature.
- there no point in finding interesting points , extract features whole image. (again) don't know how hog work either extract features whole image or first detect interesting points , extract them. (some features cannot extracted whole image though).
- judging answers in wikipedia usual approach involves calculation of local features , use blocks
group cells larger, spatially connected blocks
improve illumination etc robustness, first calculate hog on cell level , use block groug them together.
Comments
Post a Comment