Very Deep Convolutional Networks for Large-Scale Image Recognition

Karen SimonyanAndrew Zisserman

   Papers with code   Abstract  PDF

In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers... (read more)

Benchmarked Models

RANK
MODEL
REPO
CODE RESULT
PAPER RESULT
ε-REPRODUCED
BUILD
1
BN-VGG-19
75.9%
--
2
BN-VGG-16
74.3%
--
3
VGG-19
74.3%
74.5%
4
VGG-19
(batch-norm)
74.0%
--
5
BN-VGG-19b
73.9%
--
6
VGG-16
(batch-norm)
73.1%
--
7
VGG-16
73.0%
74.4%
8
BN-VGG-16b
72.8%
--
9
VGG-19
72.2%
74.5%
10
BN-VGG-13
72.2%
--
11
BN-VGG-13b
71.6%
--
12
VGG-16
71.4%
74.4%
13
VGG-13
(batch-norm)
71.4%
71.3%
14
VGG-13
71.2%
71.3%
15
BN-VGG-11
71.0%
--
16
BN-VGG-11b
70.4%
--
17
VGG-11
(batch-norm)
70.2%
70.4%
18
VGG-11
70.1%
70.4%
19
VGG-13
69.6%
71.3%
20
VGG-11
68.8%
70.4%