|
gptkbp:instance_of
|
gptkb:television_channel
|
|
gptkbp:bfsLayer
|
5
|
|
gptkbp:bfsParent
|
gptkb:Efficient_Net
|
|
gptkbp:application
|
gptkb:viewpoint
|
|
gptkbp:architectural_style
|
Image Classification
|
|
gptkbp:base
|
gptkb:Efficient_Net
|
|
gptkbp:class
|
gptkb:television_series
|
|
gptkbp:collection_size
|
1.2 million images
|
|
gptkbp:community_support
|
Active
|
|
gptkbp:developed_by
|
gptkb:Google_AI
|
|
gptkbp:features
|
gptkb:theorem
|
|
gptkbp:first_performance
|
84.4%
|
|
https://www.w3.org/2000/01/rdf-schema#label
|
Efficient Net B4
|
|
gptkbp:image
|
RGB images
|
|
gptkbp:image_stabilization
|
Resizing and normalization
|
|
gptkbp:input_output
|
(None, 1000)
380x380
|
|
gptkbp:is_a_framework_for
|
gptkb:Graphics_Processing_Unit
|
|
gptkbp:is_available_in
|
gptkb:Keras
gptkb:Py_Torch
gptkb:Tensor_Flow_Hub
|
|
gptkbp:is_cited_in
|
over 1000
|
|
gptkbp:is_evaluated_by
|
Validation set
|
|
gptkbp:is_standardized_by
|
gptkb:theorem
|
|
gptkbp:learns_move
|
gptkb:theorem
|
|
gptkbp:losses
|
Categorical Crossentropy
|
|
gptkbp:orbital_period
|
15 million
|
|
gptkbp:performance
|
FLO Ps
|
|
gptkbp:predecessor
|
gptkb:Efficient_Net_B3
|
|
gptkbp:provides_information_on
|
gptkb:Image_Net
gptkb:theorem
|
|
gptkbp:publishes
|
Efficient Net: Rethinking Model Scaling for Convolutional Neural Networks
|
|
gptkbp:rank
|
97.1%
|
|
gptkbp:reduces
|
0.3
|
|
gptkbp:related_model
|
gptkb:theorem
gptkb:Efficient_Net_B0
gptkb:Efficient_Net_B1
gptkb:Efficient_Net_B2
gptkb:Efficient_Net_B3
gptkb:Efficient_Net_B5
gptkb:Efficient_Net_B6
gptkb:Efficient_Net_B7
66 MB
|
|
gptkbp:release_year
|
gptkb:2019
|
|
gptkbp:resolution
|
gptkb:High
|
|
gptkbp:restoration
|
L2 regularization
|
|
gptkbp:scales
|
Compound scaling
|
|
gptkbp:sister_channel
|
gptkb:3
|
|
gptkbp:starts_at
|
gptkb:theorem
|
|
gptkbp:successor
|
gptkb:Efficient_Net_B5
|
|
gptkbp:throws
|
19 billion
|
|
gptkbp:training
|
0.001
several days
|
|
gptkbp:tuning
|
gptkb:Adam
gptkb:theorem
|
|
gptkbp:uses
|
Swish activation function
|