FOMO Model Based Logo Detection in Embroidery Process

Abstract

This paper presents an innovative approach to improve the accuracy and efficiency of the embroidery process by integrating machine vision systems into traditional stitching machines. The problem of maintaining high accuracy in embroidery designs is a well-known challenge, commonly addressed through manual inspection and correction. In this paper, an artificial vision technique, implemented using the Faster Objects, More Objects (FOMO) learning algorithm, which allows machines to recognize patterns and improve accuracy, is proposed. The implementation was performed on the Edge Impulse platform and using the ESP32-CAM camera, successfully recognized and classified the images, achieving a 100% recognition rate for the "First Corporal" logo, 92.3% for the "Second Corporal" logo, and 96.2% for the "Sergeant" logo. The proposed solution has significant implications for the embroidery industry, offering a more accurate and efficient alternative to traditional methods.