|
gptkbp:instanceOf
|
gptkb:convolutional_neural_network
|
|
gptkbp:application
|
transfer learning
feature extraction for other tasks
|
|
gptkbp:architecture
|
deep feedforward network
|
|
gptkbp:arXivID
|
gptkb:1409.1556
|
|
gptkbp:characteristic
|
increases depth to improve performance
uses 2x2 max pooling
uses small 3x3 convolution filters
|
|
gptkbp:developedBy
|
gptkb:University_of_Oxford
gptkb:Visual_Geometry_Group
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
VGG network
|
|
gptkbp:influenced
|
gptkb:ResNet
gptkb:DenseNet
modern CNN architectures
|
|
gptkbp:introduced
|
gptkb:Andrew_Zisserman
gptkb:Karen_Simonyan
|
|
gptkbp:introducedIn
|
2014
|
|
gptkbp:level
|
16
19
|
|
gptkbp:notablePublication
|
gptkb:Very_Deep_Convolutional_Networks_for_Large-Scale_Image_Recognition
|
|
gptkbp:notableVariant
|
gptkb:VGG16
VGG19
|
|
gptkbp:openSource
|
gptkb:TensorFlow
gptkb:Caffe
gptkb:PyTorch
|
|
gptkbp:parameter
|
138 million (VGG16)
|
|
gptkbp:powers
|
simple and uniform architecture
|
|
gptkbp:trainer
|
gptkb:ImageNet_dataset
|
|
gptkbp:usedFor
|
feature extraction
image classification
object detection
|
|
gptkbp:weakness
|
high computational cost
large number of parameters
|
|
gptkbp:bfsParent
|
gptkb:A_Neural_Algorithm_of_Artistic_Style
gptkb:Super-Resolution_Generative_Adversarial_Network
|
|
gptkbp:bfsLayer
|
7
|