Classification of Animals Using Toy Images
D. Nanditha, Namrata Manohar · 2020
Through this work, we have built a supervised, based classification system for the classification of the images of real animals. The classification model is trained using the toy images of animals to account for factors other than just the physical appearance of animals. At first, the image is preprocessed to remove noise and enhanced using the adaptive histogram equalization and median filtering techniques. In the second stage, the preprocessed toy image is segmented using the k-means clustering technique. Segmentation separates the toy animal image from the background. The third stage involves extraction of hog features from the segmented image. In the final stage, the extracted features are used to classify the image using the supervised based multi-SVM classifier to appropriate animal class. The animal image is segmented and classified based on various characteristics/features extracted from the image itself. This project also aims at achieving nearly accurate animal classification considering factors other than just the physical appearance of the animal.