Classification of Natural Scene using Convolution Neural Network
Supriya R Iyer, GECBH · International Journal of Engineering Research and · 2020
Classification is a process in which objects are recognized, differentiated and understood.Image classification is classifying image into one of the predefined classes.In conventional way, people use different computer vision techniques to extract features from images and different machine learning algorithms use these extracted features to classify the images.Nowadays, accuracy and performance of the model depends mainly on trained dataset and algorithm used.Neural networks are found to be extremely effective in classification of our data.A Convolution Neural Network concept is used.Natural scenes has objects we would ideally want computer to recognize automatically.Object is recognized on the basis of shape and textual characteristics of regions of interest.MATLAB tool is used to classify the images into their classes.There are different classes of natural scenes to be identified into their respective categories.Here we are mainly concerned with 12 different categories.Dataset containing several thousands of images of natural scene is used to train the model.Neural network model used is Alex-Net model.The histogram analysis of images are carried out at each classification of image.The accuracy of the image as well as the loss occurred in the model is also found out.The performance of the model is calculated with the help of confusion matrix which represents true value corresponding to each class.