20160909-卷积神经网络-GY分析.ppt

* CSDN 博客 知乎 GitHub QQ群 Thank you！ * * * 1. Background 2. Characteristic 3. Transfer Learning 4. MatConvNet Background In the 1960s, Hubel and Wiesel propose the local receptive field when researching the visual cortex of cat. Fukushima K. Neocognitron: A self-organizing neural network for a mechanism of pattern recognition unaffected by shift in position. Biological Cybernetics 36(4), 1980: 193–202. LeNet5 LeCun?Y,?Bottou?L,?et al.?Gradient-based?learning applied?to?document?recognition?[C]//?Proc.?IEEE,?1998:?2278-2324. AlexNet ILSVRC2012 (ImageNet large scale visual recognition challenge，ILSVRC) Krizhevsky A， Sutskever I， Hinton G E．Imagenet classification with deep convolutional neural networks［C］//The 26th Conference on Neural Information Processing Systems. 2012 20世纪90年代，各种各样的浅层机器学习模型相继被提出，例如支持向量机（SVM）、 Boosting、最大熵方法（LR）等。 * * ZF Net Zeiler M D，Fergus R．Visualizing and understanding convolutional neural networks．ECCV 2014. GoogLeNet Szegedy C, Liu W, Jia Y, et al.. Going Deeper with Convolutions. CVPR 2015. VGGNet Simonyan K, Zisserman A. A Very Deep Convolutional Networks for Large-Scale Image Recognition. ICLR 2015. ResNet He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition. CVPR 2016. * Characteristic Local receptive field * Weight sharing * Pooling * Advantages：Features are extracted automatically; Low-level features are abstracted to high-level features, which confine the semantic gap effectively. Disadvantages ：A large scale dataset; High performance hardware * Transfer Learning Oquab M, et al. Learning and Transferring Mid-Level Image Representations using Convolutional Neural Networks. CVPR 2014 * * Azizpour H, et al. Factors of Transferability for a Generic ConvNet Representation. CVPRW2015. * Gong Y, et al. Multi-Scale Orderless Pooling of Deep Convolutional Activation Features. ECCV2014 * * Ng J, et al. Exploiting Local Features from Deep Networks for Image Retrieval. CV