RELNet-An Efficient and Lightweight Classification Network
Cheng Chen, Ji Li · 2022
In the present era of information explosion, the ability of the computer to replace artificial recognition images are widely used in our life, such as face recognition, intelligent robot, information filtering and so on. Driven by the rapid development of deep learning technology and led by Convolutional Neural Network (CNN), these applications are maturer and maturer. In this paper, we designed a network for image classification on the CIFARtO, CIFARtOO and MNIST which is based on the convolutional neural network. we independently set the hyper-parameters and design the model structure to generate a convolutional neural network model with a small number of convolutional layers. Based on understanding the ResNet network structure, we got our network named RELNet. This model has certain advantages over other models with a similar number of convolutional layers, and it is an improved model that can achieve certain improvements in various evaluation indicators. In addition, this model has a faster convergence speed and fewer training parameters, which is considerable progress compared with other similar models.