Читать книгу Computational Analysis and Deep Learning for Medical Care - Группа авторов - Страница 18

Simonyan and Zisserman et al. [4] introduced VGGNet for the ImageNet Challenge in 2014. VGGNet-16 consists of 16 layers; accepts a 227 × 227 × 3 RGB image as input, by subtracting global mean from each pixel. Then, the image is fed to a series of convolutional layers (13 layers) which uses a small receptive field of 3 × 3 and uses same padding and stride is 1. Besides, AlexNet and ZFNet uses max-pooling layer after convolutional layer. VGGNet does not have max-pooling layer between two convolutional layers with 3 × 3 filters and the use of three of these layers is more effective than a receptive field of 5 × 5 and as spatial size decreases, the depth increases. The max-pooling layer uses a window of size 2 × 2 pixel and a stride of 2. It is followed by three fully connected layers; first two with 4,096 neurons and third is the output layer with 1,000 neurons, since ILSVRC classification contains 1,000 channels. Final layer is a softmax layer. The training is carried out on 4 Nvidia Titan Black GPUs for 2–3 weeks with ReLU nonlinearity activation function. The number of parameters is decreased and it is 138 million parameters (522 MB). The test set top-5 error rate during competition is 7.1%. Figure 1.4 shows the architecture of VGG-16, and Table 1.5 shows its parameters.

Table 1.4 Various parameters of ZFNet.

Layer name	Input size	Filter size	Window size	# Filters	Stride	Padding	Output size	# Feature maps	# Connections
Conv 1	224 × 224	7 × 7	-	96	2	0	110 × 110	96	14,208
Max-pooling 1	110 × 110		3 × 3	-	2	0	55 × 55	96	0
Conv 2	55 × 55	5 × 5	-	256	2	0	26 × 26	256	614,656
Max-pooling 2	26 × 26	-	3 × 3	-	2	0	13 × 13	256	0
Conv 3	13 × 13	3 × 3	-	384	1	1	13 × 13	384	885,120
Conv 4	13 × 13	3 × 3	-	384	1	1	13 × 13	384	1,327,488
Conv 5	13 × 13	3 × 3	-	256	1	1	13 × 13	256	884,992
Max-pooling 3	13 × 13	-	3 × 3	-	2	0	6 × 6	256	0
Fully connected 1	4,096 neurons								37,752,832
Fully connected 2	4,096 neurons								16,781,312
Fully connected 3	1,000 neurons								4,097,000
Softmax	1,000 classes								62,357,608 (Total)

Figure 1.4 Architecture of VGG-16.