Abstract
In this paper, we propose the new method of Multiple Object Recognition with Bag-of-Keypoints and Supervised Multidimensional Scaling.
BoK is popular method of object recognition, and use frequency histogram of local features to recognize object.
However, when BoK is applied for an image contained a number of objects, it could be difficult to output each category of objects. This is because local features picked out from some classes are mixed up in feature space. To separate local features for each class in feature space, we use Supervised Mutidimensional Scaling.
Usually, in case of area segmentation, Class Labels are assigned to each pixel of image. In this proposal, we enable to separate each class in feature space. In experimentation, we apply this method to various types of images and examine its availability.
