EfficientNet-B7

GPTKB entity

Statements (48)
Predicate Object
gptkbp:instanceOf gptkb:convolutional_neural_network
gptkbp:architecture gptkb:model
gptkbp:basedOn compound scaling method
gptkbp:developedBy gptkb:Google_AI
gptkbp:displayType 600x600
gptkbp:hasActivationFunction gptkb:Swish
gptkbp:hasDropConnect yes
gptkbp:hasDropout yes
gptkbp:hasFinalLayer Fully Connected
gptkbp:hasInputChannels 3
gptkbp:hasNormalization Batch Normalization
gptkbp:hasOutputClasses 1000 (ImageNet)
gptkbp:hasPaperAuthors gptkb:Mingxing_Tan
gptkb:Quoc_V._Le
gptkbp:hasPaperTitle gptkb:EfficientNet:_Rethinking_Model_Scaling_for_Convolutional_Neural_Networks
gptkbp:hasPaperURL https://arxiv.org/abs/1905.11946
gptkbp:hasParameterCount 66 million
gptkbp:hasPooling Global Average Pooling
gptkbp:hasSqueezeExcitation yes
gptkbp:hasTop1Accuracy 84.4% on ImageNet
gptkbp:hasTop5Accuracy 97.1% on ImageNet
gptkbp:hasVariant gptkb:EfficientNet-B0
gptkb:EfficientNet-B1
gptkb:EfficientNet-B2
gptkb:EfficientNet-B3
gptkb:EfficientNet-B4
gptkb:EfficientNet-B5
gptkb:EfficientNet-B6
EfficientNet-B8
EfficientNet-L2
https://www.w3.org/2000/01/rdf-schema#label EfficientNet-B7
gptkbp:implementedIn gptkb:TensorFlow
gptkb:PyTorch
gptkbp:improves gptkb:Inception-v4
gptkb:ResNet-152
gptkbp:inputImageSize 600x600
gptkbp:introducedIn 2019
gptkbp:license Apache 2.0
gptkbp:maximumDepth 66 layers
gptkbp:openSource yes
gptkbp:partOf EfficientNet family
gptkbp:pretrainedWeightsAvailable yes
gptkbp:publishedIn gptkb:ICML_2019
gptkbp:usedFor transfer learning
image classification
gptkbp:width 3.1x base width
gptkbp:bfsParent gptkb:EfficientNet
gptkbp:bfsLayer 6