computer vision - Problems with Image Classification -



computer vision - Problems with Image Classification -

my objective classify images 1 of few predefined categories (sportshoes, shirts, heels, watches..) catalog (and later on homecoming similar images catalog).

i using dense-sift feature extraction, representing each image using handbag of visual words , svm classification. training images taken catalog.

the problem images querying pictures taken photographic camera , these different catalog images. example, heels/sportshoes in catalog contain right shoe taken @ 1 particular angle, whereas query image contains heel , part of foot well, , angle @ photo taken can vary (deviation catalog images).

hence classification works when query(test) image image catalog (those have not used training), not images taken camera.

how proceed? problem feature vector or training info itself? if cannot alter training data, there else can use? should utilize different approach (not bag-of-words) ?

thanks

computer-vision svm sift cbir

Comments

Popular posts from this blog

php - Android app custom user registration and login with cookie using facebook sdk -

django - Access session in user model .save() -

php - .htaccess Multiple Rewrite Rules / Prioritizing -