Densely Connected Convolutional Networks

CVPR 2017 Gao HuangZhuang LiuLaurens van der MaatenKilian Q. Weinberger

Recent work has shown that convolutional networks can be substantially deeper, more accurate, and efficient to train if they contain shorter connections between layers close to the input and those close to the output. In this paper, we embrace this observation and introduce the Dense Convolutional Network (DenseNet), which connects each layer to every other layer in a feed-forward fashion... (read more)

PDF Abstract

Code


Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Image Classification CIFAR-10 DenseNet Percentage correct 96.54 # 16
Image Classification CIFAR-10 DenseNet Percentage error 3.46 # 13
Image Classification CIFAR-100 DenseNet Percentage correct 82.62 # 11
Image Classification CIFAR-100 DenseNet Percentage error 17.18 # 5
Image Classification ImageNet DenseNet-169 Top 1 Accuracy 77.92% # 67
Image Classification ImageNet DenseNet-169 Top 5 Accuracy 94.08% # 52
Image Classification ImageNet DenseNet-264 Top 1 Accuracy 79.20% # 46
Image Classification ImageNet DenseNet-264 Top 5 Accuracy 94.71% # 36
Image Classification ImageNet DenseNet-121 Top 1 Accuracy 76.39% # 78
Image Classification ImageNet DenseNet-121 Top 5 Accuracy 93.34% # 60
Image Classification ImageNet DenseNet-201 Top 1 Accuracy 78.54% # 58
Image Classification ImageNet DenseNet-201 Top 5 Accuracy 94.46% # 45
Image Classification ImageNet DenseNet-201 Number of params 20M # 1
Image Classification SVHN DenseNet Percentage error 1.59 # 9