各自深度网络特征比较
|
jackyhwei 发布于 2019-03-22 15:49 点击:次
|
|
各自深度网络特征比较
|
Name |
Acc. test |
finetuned Acc. test. |
Train time |
Forward pass time |
On disk model size |
Year |
Paper |
|
AlexNet |
93.65% |
97.98% |
35m |
3.01 ms |
227.5Mb |
2012 |
link |
|
SqeezeNet v1.1 |
92.46% |
98.87% |
~2h |
3.91 ms |
2.9Mb |
2016 |
link |
|
GoogLeNet |
94.62% |
99.58% |
50m |
11.73 ms |
41.3Mb |
2014 |
link |
|
VGG-16 |
96.51% |
99.40% |
5h20m |
15.41 ms |
537.1Mb |
2014 |
link |
|
VGG-19 |
97.42% |
99.48% |
25h50m |
19.23 ms |
558.3Mb |
2014 |
link |
|
Network-In-Network |
93.65% |
98.49% |
~2h |
3.17 ms |
26.3Mb |
2014 |
link |
|
ResNet-50 |
95.84% |
99.52% |
18h |
24.91 ms |
94.3Mb |
2015 |
link |
|
ResNet-101 |
96.39% |
99.48% |
1d 20h |
40.95 ms |
170.5Mb |
2015 |
link |
(lien0906) |
------分隔线----------------------------
