PolygoNet: Leveraging Simplified Polygonal Representation for Effective Image Classification

TL;DR

This work proposes an efficient approach that leverages polygonal representations of images using dominant points or contour coordinates that significantly reduces computational requirements, accelerates training, and conserves resources making it suitable for real time and resource constrained applications.

Abstract

Deep learning models have achieved significant success in various image related tasks. However, they often encounter challenges related to computational complexity and overfitting. In this paper, we propose an efficient approach that leverages polygonal representations of images using dominant points or contour coordinates. By transforming input images into these compact forms, our method significantly reduces computational requirements, accelerates training, and conserves resources making it suitable for real time and resource constrained applications. These representations inherently capture essential image features while filtering noise, providing a natural regularization effect that mitigates overfitting. The resulting lightweight models achieve performance comparable to state of the art methods using full resolution images while enabling deployment on edge devices. Extensive experiments on benchmark datasets validate the effectiveness of our approach in reducing complexity, improving generalization, and facilitating edge computing applications. This work demonstrates the potential of polygonal representations in advancing efficient and scalable deep learning solutions for real world scenarios. The code for the experiments of the paper is provided in https://github.com/salimkhazem/PolygoNet.

Figures (3)

• Figure 1: The first step in our shape encoding process involves applying thresholding to the image to segment the object from the background. This is followed by extracting the contours using one of the methods detailed in Section \ref{['sec:contours_extraction']}. The number of contour points obtained varies depending on the extraction method used and the complexity of the shape.
• Figure 2: The initial step in encoding a shape begins with applying thresholding to the image, followed by contour extraction, and finally applying the Modified Adaptive Tangential Cover (MATC) algorithm to compute the dominant points. The number of dominant points is variable and depends on the complexity of the shape.
• Figure 3: PolygoNet pipeline. The input colored image is converted to grayscale before being thresholded with Otsu. The dominant points are extracted using the MATC approach from the extracted contour. This variable size sequence of dominant points is then processed for classification by PolygoNet.