Research on Bottle Defect Detection Based on Improved FCOS

Abstract

In order to solve the problem of slow speed and low efficiency of product defect detection, this paper proposes an improved Fully Convolutional One-Stage Object Detection (FCOS) algorithm for detection, and compares it with several currently widely used target detection algorithms. Through experimental verification, the improved Fully Convolutional One-Stage Object Detection (FCOS) has improved the detection effect by 6.9% compared with the original Fully Convolutional One-Stage Object Detection (FCOS) algorithm, and the applicability of the improved Fully Convolutional One-Stage Object Detection algorithm is verified through experiments.