PolygoNet: Leveraging Polygonal Contours for Efficient Image Classification with deep neural networks
Abstract: In recent years, deep learning models have demonstrated remarkable capabilities in various image-related tasks, yet they are often plagued by computational complexity and susceptibility to overfitting. In this paper, we propose a novel approach that leverages efficient polygon representation through dominant points for the input images to address these challenges for image classification tasks. Our method focuses on transforming input images into polygon representations, which are subsequently utilized for training deep neural networks. The key contribution lies in the use of theses dominant points, which offer a concise and flexible representation of images. By transforming images into dominant points, we significantly reduce the computational burden associated with processing large image datasets. This reduction in calculation not only accelerates the training process but also conserves computational resources, making our approach particularly appealing for real-time applications and resource-constrained environments. We validate our approach through extensive experiments on benchmark datasets, showcasing its effectiveness in reducing computation. The experimental results demonstrate that our method achieves state-of-the-art performance across various image classification tasks, underscoring its potential on standard configuration and edge computing configuration.
Submission Length: Regular submission (no more than 12 pages of main content)
Assigned Action Editor: ~Yunhe_Wang1
Submission Number: 2958
Loading