Abstract

Visual image classification plays an important role in computer vision and pattern recognition. In this paper, a new random forests method called metric forests is suggested. This method takes the distribution of datasets (including the original dataset and bootstrapped ones) into full consideration. The proposed method exploits the distribution similarity between the original dataset and the bootstrapped datasets. For each bootstrapped dataset, a metric decision tree is built based on Gaussian mixture model. The metric decision tree learned from bootstrapped dataset with a low or high similarity index is given small weight when voting, vice versa. The contribution of the proposed method is originated from that the dataset with low similarity may not represent the original dataset very well while the high one with a big chance to overfit. To evaluate the proposed metric forests method, extensive of experiments was conducted for visual image classification including texture image classification, flower image classification and food image classification. The experimental results validated the superiority of the proposed metric forests on the ALOT, Flower-102 and Food-101 datasets.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call