Improved Image Classification Accuracy by Convolutional Neural Networks
Litao Liu · 2021
Along with the development of mobile Internet, images are increasingly used as a more effective information carrier, and deep learning methods have a great efficiency advantage over traditional algorithms in dealing with image classification problems. This paper presents a simple and efficient deep learning method to train a convolutional neural network (CNN) for image classification on a large-scale supervised fashion NBIST image dataset. We first preprocessed the dataset, after which it was fed into the neural network for a minimum of five epochs of training. In the end, the accuracy rate can reach more than 83.6%. This CNN model has a greater efficiency advantage over traditional feature description and detection algorithms and is more suitable for today's image classification tasks.