人工檢查的效率低無法趕上快速擴張的製造規模，更遑論人工檢查的高昂成本。當今計算機視覺蓬勃發展，在重複的大批量製造環境中，數據非常豐富，適合模型開發，機器學習模型透過模仿人類腦部神經元活動，對資料進行分析與推理，從而顯著降低成本，透過遠程檢查產品避免接觸物料，適合於智能製造領域中廣泛實施。因此，使用卷積神經網路（Convolutional Neural Network, CNN）是一個好的選擇。然而，最大的問題之一是標記數據的高成本，因此我們使用遷移學習，主要原理是應用各種任務共有的通用原理與抽象結構，以加速電腦的學習速度。本研究探討三種不同的方法:全層凍結的遷移學習(Transfer learning)、結合遷移學習與遷移學習結合隨機森林(Transfer learning + Random forest)與結合遷移學習與微調(Fine-tune)。通過各方法於圖像識別之表現，以比較三種不同模式。除了準確率外，本研究還使用精確率(Precision rate)、召回率(Recall rate)和F1分數作為整體衡量指標。經由數據分析的結果，我們觀察到在每種方法中，單獨使用遷移學習並不能獲得良好的性能。遷移學習是一種合適的中間工具，可幫助我們提取特徵並將非結構化細節轉換為向量，但它並不適合直接作為最終的分類器，它需要與其他方法結合以提高其準確率。

The low efficiency of manual inspection cannot keep up with the rapidly expanding manufacturing scale, not to mention the high cost of manual inspection.
Computer vision is booming nowadays. In the repeated mass manufacturing environment, the large amount of data is suitable for model development. Machine learning model works by imitating the interaction of brain neurons, thus significantly reducing costs. By remotely checking products to avoid contacting the materials, it is suitable for widespread implementation in the field of intelligent manufacturing. Using Convolutional Neural Network (CNN) is a suitable choice. However, one of the biggest problems is the high cost of labeling data. We adopt transfer learning to overcome the issue. The main concept of transfer learning is to apply common principles and structures to various tasks to accelerate the learning speed of computers. This study explores three different methods: full-layer frozen transfer learning, combination of transfer learning and feature extraction, and combination of transfer learning and fine-tuning. Evaluating each mode’s performance by image recognition. In addition to the accuracy rate, this study also uses the precision rate, recall rate and F1 score as the overall measurement indicators. According to the result of data analysis, we observe that in each method, using "transfer learning" alone does not achieve good performance.
Transfer learning is a suitable intermediate tool that can help us extract features and convert unstructured details into vectors, but it is not suitable for direct use as the final classifier. It needs to be combined with other methods to improve its accuracy.

1. Breiman, L. (2001). Random Forests, Machine Learning, 45(1), 5-32
2. Cao, G., Wang, S., Wei, B., Yin, Y., & Yang, G. (2013). A hybrid CNN-RF method for electron microscopy images segmentation. Tissue Engineering, J. Biomim Biomater Tissue Eng, 18, 2.
3. Dietterich, T.G. (2002). Ensemble Learning. Machine Learning, 1-15
4. Freund, Y., & Schapire, R. E. (1996). Experiments with a new boosting algorithm. In icml, 96, 148-156.
5. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 770-778
6. Kim, J., Kwon Lee, J., & Mu Lee, K. (2016). Accurate image super-resolution using very deep convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, 1646-1654
7. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems, 1097-1105
8. Kumar, A., Lyndon, D., Kim, J., & Feng, D. (2016). Subfigure and Multi-Label Classification using a Fine-Tuned Convolutional Neural Network. In CLEF (Working Notes), 318-321
9. LeCun, Y., Haffner, P., Bottou, L., & Bengio, Y. (1999). Object recognition with gradient-based learning. In Shape, contour and grouping in computer vision, 319-345
10. Lim, S., Lee, K., Byeon, O., & Kim, T. (2001). Efficient iris recognition through improvement of feature vector and classifier. ETRI journal, 23(2), 61-70.
11. Pan, S. J., & Yang, Q. (2009). A survey on transfer learning. IEEE Transactions on knowledge and data engineering, 22(10), 1345-1359.
12. Rosenstein, M. T., Marx, Z., Kaelbling, L. P., & Dietterich, T. G. (2005). To transfer or not to transfer. In NIPS 2005 workshop on transfer learning, 898,1-4
13. Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. nature, 323(6088), 533-536.
14. Schroff, F., Kalenichenko, D., & Philbin, J. (2015). Facenet: A unified embedding for face recognition and clustering. In Proceedings of the IEEE conference on computer vision and pattern recognition, 815-823
15. Shin, H. C., Roth, H. R., Gao, M., Lu, L., Xu, Z., Nogues, I., Yao, J., Mollura, D., Summers, R. M. (2016). Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning. IEEE transactions on medical imaging, 35(5), 1285-1298.
16. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition, 2818-2826
17. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., Rabinovich, A. (2015). Going deeper with convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition, 1-9
18. Wang, H. (2018). ANET: Automated Optical Inspection Network.
19. Yin, X., Han, J., Yang, J., & Yu, P. S. (2006). Efficient classification across multiple database relations: A crossmine approach. IEEE Transactions on Knowledge and Data Engineering, 18(6), 770-783.
20.Zhang, H., Liu, D., & Xiong, Z. (2017). Cnn-based text image super-resolution tailored for ocr. In 2017 IEEE Visual Communications and Image Processing