Pré-Publication, Document De Travail Année : 2022

Real Time Automatic Seal Defect Detection Based on Deep Learning Method for Mono-material Flexible Packaging in Manufacturing

Résumé

Automatic seal defect detection in manufacturing represents a technical challenge in computer vision and quality control. In this paper, we have developed a computer vision system based on deep learning models for real-time automatic detection of sealing defects in mono-material flexible packaging. Different deep learning approaches have been studied, to understand their effectiveness and inefficiencies to real time automatic seal defect detection. To reduce the computational and memory requirements of deep learning networks, we have used the classical pruning and quantization methods based on NVIDIA's Cutting-Edge GPUs for deep learning models. The experimental results show that the proposed method can achieve an accuracy rate of 98.7\%. precision (80.69\%) and recall (88.89\%). These results illustrate the effectiveness of our deep learning models for seal defect detection and quality control.

Fichier non déposé

Dates et versions

hal-05034330 , version 1 (15-04-2025)

Identifiants

  • HAL Id : hal-05034330 , version 1

Citer

Tajmout Ilyas, Hamid Ladjal, Fayez-Shakil Ahmed, Frédéric Roumanet, Hassan Hammouri. Real Time Automatic Seal Defect Detection Based on Deep Learning Method for Mono-material Flexible Packaging in Manufacturing. 2022. ⟨hal-05034330⟩
141 Consultations
0 Téléchargements

Partager

  • More