commit files to HF hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,33 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# vgg11
|
| 2 |
+
Implementation of VGG proposed in [Very Deep Convolutional Networks For
|
| 3 |
+
Large-Scale Image Recognition](https://arxiv.org/pdf/1409.1556.pdf)
|
| 4 |
+
|
| 5 |
+
``` python
|
| 6 |
+
VGG.vgg11()
|
| 7 |
+
VGG.vgg13()
|
| 8 |
+
VGG.vgg16()
|
| 9 |
+
VGG.vgg19()
|
| 10 |
+
VGG.vgg11_bn()
|
| 11 |
+
VGG.vgg13_bn()
|
| 12 |
+
VGG.vgg16_bn()
|
| 13 |
+
VGG.vgg19_bn()
|
| 14 |
+
```
|
| 15 |
+
|
| 16 |
+
Please be aware that the [bn]{.title-ref} models uses BatchNorm but
|
| 17 |
+
they are very old and people back then don\'t know the bias is
|
| 18 |
+
superfluous in a conv followed by a batchnorm.
|
| 19 |
+
|
| 20 |
+
Examples:
|
| 21 |
+
|
| 22 |
+
``` python
|
| 23 |
+
# change activation
|
| 24 |
+
VGG.vgg11(activation = nn.SELU)
|
| 25 |
+
# change number of classes (default is 1000 )
|
| 26 |
+
VGG.vgg11(n_classes=100)
|
| 27 |
+
# pass a different block
|
| 28 |
+
from nn.models.classification.senet import SENetBasicBlock
|
| 29 |
+
VGG.vgg11(block=SENetBasicBlock)
|
| 30 |
+
# store the features tensor after every block
|
| 31 |
+
```
|
| 32 |
+
|
| 33 |
+
|