Design And Implementation Of Image Classification Model Based On Deep Convolution Neural Network

Every single day numerous images are produced,which creates the necessity to classify them so that accessibility is easier and faster?Artificial image and simple attributes(such as date,etc.)image classification methods are unable to classify the rapidly increasing images efficiently.We trained a large,deep convolutional neural network to classify the 1.2 million high-resolution images in the ImageNet LSVRC-2010 contest into the 1000 different classes.The neural network,which has 60 million parameters and 650,000 neurons,consists of five convolutional layers,some of which are followed by max-pooling layers,and three fully-connected layers with a final 1000-way softmax.To improve the training speed,we deploy the image classification model on a high-performance GPU through a general-purpose computing platform and change the activation function in neurons at all layers of the model from Sigmoid function to ReLUs.To prevent the model from over-fitting,we apply the affine transformation method to augment the image data and "Dropout" method in the network structure.In addition,to prevent the model from ICS during training,we normalized the images during the training.The entire classification model is implemented on Ubuntu system,we use a convolutional neural network framework named Caffe to deploy and train the model.It cost about 7 days to finish the training.On the validation data,we achieved top-1 and top-5 error rates of 36.7%and 18.2%which is considerably better than the previous state-of-the-art.The trained model can be called directly from the command line and applied to the image classification scene in life.