Defects Detection Based on Deep Learning and Transfer Learning

Defect detection is an important step in the field of industrial production. Through the study of deep learning and transfer learning, this paper proposes a method of defect detection based on deep learning and transfer learning. Our method firstly establishes Deep Belief Networks and trains it according to the source domain sample feature, in order to obtain the weights of the network according to source domain samples. Then, the structure and parameters of the source domain DBN is transferred to the target domain and target domain samples are used to fine-tune the network parameters to get the mapping relationship between the target domain training sample and defect-free template. Finally, the defects of testing samples will be detected by compared with the reconstruction image. This method not only can make full use of the advantages of DBN, also can solve over-fitting in DBN network training through parameters transfer learning. These experiments show that DBN is a successful approach in the high-dimensional-feature-space information extraction task, which can perfectly establishes the mapping relationship, and can quickly detect defects with a high accuracy.

[1]  Tara N. Sainath,et al.  Deep Neural Networks for Acoustic Modeling in Speech Recognition , 2012 .

[2]  Geoffrey E. Hinton,et al.  Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[3]  Yoshua Bengio,et al.  Unsupervised and Transfer Learning Challenge: a Deep Learning Approach , 2011, ICML Unsupervised and Transfer Learning.

[4]  Songhwai Oh,et al.  Human behavior prediction for smart homes using deep learning , 2013, 2013 IEEE RO-MAN.

[5]  Geoffrey E. Hinton.  Training Products of Experts by Minimizing Contrastive Divergence , 2002, Neural Computation.

[6]  Wang Xian,et al.  Solar Cells Surface Defects Detection Based on Deep Learning , 2014 .

[7]  Yoshua Bengio,et al.  Deep Learning of Representations for Unsupervised and Transfer Learning , 2011, ICML Unsupervised and Transfer Learning.

[8]  Yoshua Bengio,et al.  Extracting and composing robust features with denoising autoencoders , 2008, ICML '08.

[9]  Geoffrey E. Hinton,et al.  Autoencoders, Minimum Description Length and Helmholtz Free Energy , 1993, NIPS.

[10]  Yee Whye Teh,et al.  A Fast Learning Algorithm for Deep Belief Nets , 2006, Neural Computation.

[11]  Dong Yu,et al.  Context-Dependent Pre-Trained Deep Neural Networks for Large-Vocabulary Speech Recognition , 2012, IEEE Transactions on Audio, Speech, and Language Processing.

[12]  Hu Rong.  The Automatic Inspection System of Capsule Product Based on BP Neural Network , 2002 .

[13]  Paul Smolensky,et al.  Information processing in dynamical systems: foundations of harmony theory , 1986 .