Fast image classification by boosting fuzzy classifier
Anup Badhe, Samuel Hsiao-Heng Chang · 2016
Traditional Bag of Visual words (BoW) algorithms classify images using feature descriptors with rules to classify the features in an image. In this research to detect everyday objects, we propose to start with a set of labeled positive image samples and scale them down so that most objects in the images were represented as basic geometric shapes or approximations just as the human eye would interpret a very distant object. These geometric shapes are assigned scores that are fuzzy scores and now stored at the top of a tree or the initial starting point for the classification process. In the subsequent iterations, the images are made bigger and more composite geometric shapes were extracted since the clarity of the images was enhanced.