透過您的圖書館登入
IP:3.149.251.154
  • 期刊

基於深層殘差卷積網路應用於紡織品數位之對花檢驗

Deep Residual Network-Based Automated Fabric Inspection

摘要


因應智慧製造的快速發展,利用人工智慧解決產業問題日益增加。本研究採用監督式深層捲積神經網路,學習紡織中破洞、漏紗與斷經特徵,並利用遷移式學習與微調(Fine-tuning)提高訓練速度與準確度,藉由訓練後神經網路達到自動化檢測之目標,於2080Ti硬體架構下本研究使用之神經網路預測一張600x600圖片時間為0.189秒。

並列摘要


In response to the rapid development of smart manufacturing, the use of artificial intelligence to solve industrial problems is increasing. This research uses a supervised deep convolutional neural network to learn the characteristics of holes, missing yarn and broken warp in textiles. Moreover using transfer learning and fine-tuning to improve training speed and accuracy. After the neural network achieves the goal of automatic detection, this network predicts that it will take 0.189 seconds for 600*600 images on a 2080Ti GPU.

延伸閱讀