https://github.com/szagoruyko/fastrcnn-models.torch
Fast-RCNN models in Torch-7 format
https://github.com/szagoruyko/fastrcnn-models.torch
Last synced: about 2 months ago
JSON representation
Fast-RCNN models in Torch-7 format
- Host: GitHub
- URL: https://github.com/szagoruyko/fastrcnn-models.torch
- Owner: szagoruyko
- Created: 2016-01-18T18:05:55.000Z (over 9 years ago)
- Default Branch: master
- Last Pushed: 2016-01-18T18:06:31.000Z (over 9 years ago)
- Last Synced: 2025-02-09T18:17:42.422Z (3 months ago)
- Size: 1000 Bytes
- Stars: 18
- Watchers: 7
- Forks: 4
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Fast-RCNN models in Torch-7 format
AlexNet and VGG-16 pretrained on ImageNet and finetuned on VOC-2007 converted to
Torch-7 format.
Original models can be found here https://github.com/rbgirshick/fast-rcnn/tree/master/data
Files:
1. `imagenet/alexnet/features.t7` and `imagenet/alexnet/top.t7`: AlexNet pretrained on ImageNet
2. `imagenet/vgg/features.t7` and `imagenet/vgg/top.t7`: VGG-16 pretrained on ImageNet
3. `caffenet_fast_rcnn_iter_40000.t7` AlexNet pretrained on ImageNet and finetuned on VOC2007
4. `vgg16_fast_rcnn_iter_40000.t7` VGG-16 pretrained on ImageNet and finetuned on VOC2007
## Download link
The files are in Yandex cloud storage:
https://yadi.sk/d/R6K_3Dk9nHXn7
## Requirements
```
luarocks install cudnn
luarocks install inn
```
## Loading
Models were cleaned with `nn.Module.clearState` and `gradWeight`, `gradBias`
removed so don't forget to run `unpack()` to restore them if needed.
```lua
require 'cudnn'
require 'inn'
net = torch.load'caffenet_fast_rcnn_iter_40000.t7':unpack()
```
## Model printout
### Finetuned
#### AlexNet
```
nn.Sequential {
(1): nn.ParallelTable {
input
|`-> (1): nn.Sequential {
| (1): cudnn.SpatialConvolution(3 -> 96, 11x11, 4,4, 5,5)
| (2): cudnn.ReLU
| (3): cudnn.SpatialMaxPooling(3,3,2,2,1,1)
| (4): cudnn.SpatialCrossMapLRN
| (5): cudnn.SpatialConvolution(96 -> 256, 5x5, 1,1, 2,2)
| (6): cudnn.ReLU
| (7): cudnn.SpatialMaxPooling(3,3,2,2,1,1)
| (8): cudnn.SpatialCrossMapLRN
| (9): cudnn.SpatialConvolution(256 -> 384, 3x3, 1,1, 1,1)
| (10): cudnn.ReLU
| (11): cudnn.SpatialConvolution(384 -> 384, 3x3, 1,1, 1,1)
| (12): cudnn.ReLU
| (13): cudnn.SpatialConvolution(384 -> 256, 3x3, 1,1, 1,1)
| (14): cudnn.ReLU
| }
|`-> (2): nn.Identity
... -> output
}
(2): inn.ROIPooling
(3): nn.View
(4): nn.Linear(9216 -> 4096)
(5): cudnn.ReLU
(6): nn.Dropout(0.500000)
(7): nn.Linear(4096 -> 4096)
(8): cudnn.ReLU
(9): nn.Dropout(0.500000)
(10): nn.ConcatTable {
input
|`-> (1): nn.Linear(4096 -> 21)
|`-> (2): nn.Linear(4096 -> 84)
... -> output
}
}
```
### VGG-16
```
nn.Sequential {
(1): nn.ParallelTable {
input
|`-> (1): nn.Sequential {
| (1): cudnn.SpatialConvolution(3 -> 64, 3x3, 1,1, 1,1)
| (2): cudnn.ReLU
| (3): cudnn.SpatialConvolution(64 -> 64, 3x3, 1,1, 1,1)
| (4): cudnn.ReLU
| (5): cudnn.SpatialMaxPooling(2,2,2,2)
| (6): cudnn.SpatialConvolution(64 -> 128, 3x3, 1,1, 1,1)
| (7): cudnn.ReLU
| (8): cudnn.SpatialConvolution(128 -> 128, 3x3, 1,1, 1,1)
| (9): cudnn.ReLU
| (10): cudnn.SpatialMaxPooling(2,2,2,2)
| (11): cudnn.SpatialConvolution(128 -> 256, 3x3, 1,1, 1,1)
| (12): cudnn.ReLU
| (13): cudnn.SpatialConvolution(256 -> 256, 3x3, 1,1, 1,1)
| (14): cudnn.ReLU
| (15): cudnn.SpatialConvolution(256 -> 256, 3x3, 1,1, 1,1)
| (16): cudnn.ReLU
| (17): cudnn.SpatialMaxPooling(2,2,2,2)
| (18): cudnn.SpatialConvolution(256 -> 512, 3x3, 1,1, 1,1)
| (19): cudnn.ReLU
| (20): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
| (21): cudnn.ReLU
| (22): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
| (23): cudnn.ReLU
| (24): cudnn.SpatialMaxPooling(2,2,2,2)
| (25): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
| (26): cudnn.ReLU
| (27): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
| (28): cudnn.ReLU
| (29): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
| (30): cudnn.ReLU
| }
|`-> (2): nn.Identity
... -> output
}
(2): inn.ROIPooling
(3): nn.View
(4): nn.Linear(25088 -> 4096)
(5): cudnn.ReLU
(6): nn.Dropout(0.500000)
(7): nn.Linear(4096 -> 4096)
(8): cudnn.ReLU
(9): nn.Dropout(0.500000)
(10): nn.ConcatTable {
input
|`-> (1): nn.Linear(4096 -> 21)
|`-> (2): nn.Linear(4096 -> 84)
... -> output
}
}
```
### Imagenet pretrained
#### Alexnet features
```
nn.Sequential {
(1): cudnn.SpatialConvolution(3 -> 96, 11x11, 4,4, 5,5)
(2): cudnn.ReLU
(3): cudnn.SpatialMaxPooling(3,3,2,2,1,1)
(4): cudnn.SpatialCrossMapLRN
(5): cudnn.SpatialConvolution(96 -> 256, 5x5, 1,1, 2,2)
(6): cudnn.ReLU
(7): cudnn.SpatialMaxPooling(3,3,2,2,1,1)
(8): cudnn.SpatialCrossMapLRN
(9): cudnn.SpatialConvolution(256 -> 384, 3x3, 1,1, 1,1)
(10): cudnn.ReLU
(11): cudnn.SpatialConvolution(384 -> 384, 3x3, 1,1, 1,1)
(12): cudnn.ReLU
(13): cudnn.SpatialConvolution(384 -> 256, 3x3, 1,1, 1,1)
(14): cudnn.ReLU
}
```
#### Alexnet top
```
nn.Sequential {
(1): nn.Linear(9216 -> 4096)
(2): cudnn.ReLU
(3): nn.Dropout(0.500000)
(4): nn.Linear(4096 -> 4096)
(5): cudnn.ReLU
(6): nn.Dropout(0.500000)
}
```
#### VGG-16 features
```
nn.Sequential {
(1): cudnn.SpatialConvolution(3 -> 64, 3x3, 1,1, 1,1)
(2): cudnn.ReLU
(3): cudnn.SpatialConvolution(64 -> 64, 3x3, 1,1, 1,1)
(4): cudnn.ReLU
(5): cudnn.SpatialMaxPooling(2,2,2,2)
(6): cudnn.SpatialConvolution(64 -> 128, 3x3, 1,1, 1,1)
(7): cudnn.ReLU
(8): cudnn.SpatialConvolution(128 -> 128, 3x3, 1,1, 1,1)
(9): cudnn.ReLU
(10): cudnn.SpatialMaxPooling(2,2,2,2)
(11): cudnn.SpatialConvolution(128 -> 256, 3x3, 1,1, 1,1)
(12): cudnn.ReLU
(13): cudnn.SpatialConvolution(256 -> 256, 3x3, 1,1, 1,1)
(14): cudnn.ReLU
(15): cudnn.SpatialConvolution(256 -> 256, 3x3, 1,1, 1,1)
(16): cudnn.ReLU
(17): cudnn.SpatialMaxPooling(2,2,2,2)
(18): cudnn.SpatialConvolution(256 -> 512, 3x3, 1,1, 1,1)
(19): cudnn.ReLU
(20): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
(21): cudnn.ReLU
(22): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
(23): cudnn.ReLU
(24): cudnn.SpatialMaxPooling(2,2,2,2)
(25): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
(26): cudnn.ReLU
(27): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
(28): cudnn.ReLU
(29): cudnn.SpatialConvolution(512 -> 512, 3x3, 1,1, 1,1)
(30): cudnn.ReLU
}
```
#### VGG-16 top
```
nn.Sequential {
[input -> (1) -> (2) -> (3) -> (4) -> (5) -> (6) -> output]
(1): nn.Linear(25088 -> 4096)
(2): cudnn.ReLU
(3): nn.Dropout(0.500000)
(4): nn.Linear(4096 -> 4096)
(5): cudnn.ReLU
(6): nn.Dropout(0.500000)
}
```
## Credits
Converted by @szagoruyko 2015