Why CNN for Image Convolution.... Convolutional neural network Architecture,Pooling.... What to learn Feature extraction ,.. a bit of history AlexNet,VGG,ResNet,... Hardware and software GPU,Tensorflow,Pytorch,... Hangzhou Dianzi Universi的y杭州电子科技大学 School of Computer Science and Technology计算机学院周文库
Hangzhou Dianzi University 杭州电子科技大学 School of Computer Science and Technology 计算机学院 周文晖 Why CNN for Image ？ Convolution，… What to learn？ Feature extraction ，… Convolutional neural network Architecture, Pooling, … a bit of history AlexNet, VGG, ResNet, … ？ Hardware and software GPU, Tensorflow, Pytorch, …
Why CNN for Image Convolution,... Convolutional neural network Architecture,Pooling.... What to learn Feature extraction ,.. a bit of history AlexNet,VGG,ResNet,... Hardware and software GPU,Tensorflow,Pytorch,... Hangzhou Dianzi Universi的y杭州电子科技大学 School of Computer Science and Technology计算机学院周文库
Hangzhou Dianzi University 杭州电子科技大学 School of Computer Science and Technology 计算机学院 周文晖 Why CNN for Image ？ Convolution，… What to learn？ Feature extraction ，… Convolutional neural network Architecture, Pooling, … a bit of history AlexNet, VGG, ResNet, … ？ Hardware and software GPU, Tensorflow, Pytorch, …
Why CNN for Image? How to use spatial structure Input:2D image. Idea:connect patches of input to Array of pixel values neurons in hidden layer: Neuron connected to region of input Only"sees"these values. Hangchou①ianzi Universi的杭州电子科技大学 School of Computer Science and Tecfnology计算机学院周文库
Hangzhou Dianzi University 杭州电子科技大学 School of Computer Science and Technology 计算机学院 周文晖 How to use spatial structure Why CNN for Image？ Input: 2D image. Array of pixel values
Why CNN for Image? How to use spatial structure Connect patch in input layer to a single neuron in subsequent layer. Use a sliding window to define connections. How can we weight the patch to detect particular features? Hangzhou Dianzi Universi的抗州电子科技大学 School of Computer Science and Tecfnology计算机学院周文库
Hangzhou Dianzi University 杭州电子科技大学 School of Computer Science and Technology 计算机学院 周文晖 Connect patch in input layer to a single neuron in subsequent layer. Use a sliding window to define connections. How can we weight the patch to detect particular features? How to use spatial structure Why CNN for Image？
Why CNN for Image? Convolution Layer For a neuron in hidden layer: -Take inputs from patch the neuron“sees” -Compute weighted sum -Apply bias 4x4 filter:matrix applying a window of weights of weights ij Oj xi+pjtq+b computing linear combinations activating with non-linear function for neuron(p.q)in hidden layer Hangchou①ianzi Universi的y抗州电子科技大学 School of Computer Science and Technology计算机学院周文库
Hangzhou Dianzi University 杭州电子科技大学 School of Computer Science and Technology 计算机学院 周文晖 Convolution Layer Why CNN for Image？ For a neuron in hidden layer: - Take inputs from patch the neuron “sees” - Compute weighted sum - Apply bias applying a window of weights computing linear combinations activating with non-linear function