Image Classification using Deep Learning: A Comparative Study of VGG-16, InceptionV3 and EfficientNet B7 Models
Shivam Aggarwal, Ashok Kumar Sahoo, Chetan Bansal, Pradeepta Kumar Sarangi
Abstract
Image classification is the process of identification and classification of an input image or visual from a predetermined set of labeled images. This work comes under computer vision and machine learning. This work is dedicated towards image classification using deep learning by comparing various classification models based on their architecture performance and accuracy. Convolutional Neural Network (CNN) is a type of Artificial Neural Network (ANN) that learns the structural representation of an image and make predictions about its contents. The main layers of the model, known as the Convolutional layers. It detects and analyzes specific patterns in the image using filters. In this research development, image classification is done on several images of butterflies and spiders. The data set has been acquired from the ImageNet dataset. The technological progress and innovation include the observations of different CNN models to find out which one gives the most accurate outputs, i.e., identifying the type of insect in the input image. The focus has been on image recognition, data set size, classification techniques, and recognition accuracy. Three machine learning models such as VGG-16, InceptionV3 and EfficientNetB7 have been implemented in this work and the accuracies observed from the models are 97.67 %, 97.2 % and 99 % respectively. From the results it can be seen that EfficientNetB7 was found to be the best among the three models.