LiteNet-CNN: A novel efficient lightweight image classification network

LiteNet-CNN is a novel convolutional neural network architecture specifically designed for resource-constrained environments such as mobile devices and retail settings. The paper highlights the increasing demand for efficient neural networks capable of delivering strong performance on devices with limited computational power. In response to this challenge, LiteNet-CNN, a cutting-edge neural network architecture, is introduced. LiteNet-CNN significantly reduces the number of trainable parameters by approximately 78.56% compared to AlexNet and 6.34% relative to LeNet, resulting in a network with a remarkably small footprint of 41,610 parameters. Despite its lightweight design, LiteNet-CNN demonstrates impressive performance in image classification, achieving a precision of 98.42%, a recall of 97.84%, and a test accuracy of 88.66%. The model’s training accuracy on the Fashion MNIST dataset is reported as 90.04%. Due to its efficiency and compactness, LiteNet-CNN holds significant potential for deployment in various applications, including product scanning at checkout, enhancing customer experiences in fitting rooms, and accelerating image recognition on mobile devices.