Densely Connected Convolutional Networks

GPTKB entity

Statements (42)
Predicate Object
gptkbp:instanceOf deep learning architecture
gptkbp:abbreviation gptkb:DenseNet
gptkbp:application image classification
object detection
semantic segmentation
gptkbp:architecture gptkb:feedforward_neural_network
gptkbp:arXivID arXiv:1608.06993
gptkbp:bench gptkb:SVHN
gptkb:CIFAR-10
gptkb:CIFAR-100
gptkb:ImageNet
gptkbp:citation highly cited
gptkbp:contribution reduces number of parameters
alleviates vanishing-gradient problem
improves information flow and gradient propagation
introduces dense connectivity between layers
gptkbp:features feature reuse
improved parameter efficiency
compact model size
each layer receives input from all previous layers
https://www.w3.org/2000/01/rdf-schema#label Densely Connected Convolutional Networks
gptkbp:improves ResNet on several benchmarks
gptkbp:influenced subsequent neural network architectures
gptkbp:notablePublication gptkb:Densely_Connected_Convolutional_Networks
gptkbp:openSource gptkb:TensorFlow
gptkb:Keras
gptkb:PyTorch
gptkbp:proposedBy gptkb:Laurens_van_der_Maaten
gptkb:Gao_Huang
gptkb:Kilian_Q._Weinberger
gptkb:Zhuang_Liu
gptkbp:publicationYear 2017
gptkbp:publishedIn gptkb:CVPR_2017
gptkbp:relatedTo gptkb:ResNet
Convolutional Neural Networks
gptkbp:uses batch normalization
convolutional layers
bottleneck layers
ReLU activation
transition layers
gptkbp:bfsParent gptkb:DenseNet
gptkbp:bfsLayer 6