PROJECT TITLE :

Random Forest Classifier for Zero-Shot Learning Based on Relative Attribute - 2017

ABSTRACT:

For the zero-shot image classification with relative attributes (RAs), the traditional method requires that not only all seen and unseen pictures obey Gaussian distribution, but conjointly the classifications on testing samples are made by maximum chance estimation. We have a tendency to thus propose a completely unique zero-shot image classifier known as random forest based mostly on relative attribute. Initial, based on the ordered and unordered pairs of pictures from the seen classes, the idea of ranking support vector machine is used to learn ranking functions for attributes. Then, in step with the relative relationship between seen and unseen classes, the RA ranking-score model per attribute for each unseen image is made, where the acceptable seen categories are automatically selected to participate in the modeling method. Within the third step, the random forest classifier is trained based on the RA ranking legion attributes for all seen and unseen pictures. Finally, the class labels of testing images can be predicted via the trained RF. Experiments on Outdoor Scene Recognition, Pub Fig, and Shoes information sets show that our proposed method is superior to many state-of-the-art methods in terms of classification capability for zero-shot learning problems.