-
Notifications
You must be signed in to change notification settings - Fork 20
Expand file tree
/
Copy pathsurgery-instructions.txt
More file actions
14 lines (10 loc) · 993 Bytes
/
surgery-instructions.txt
File metadata and controls
14 lines (10 loc) · 993 Bytes
1
2
3
4
5
6
7
8
9
10
11
12
13
14
1. Download the pretrained models and the deploy.protxt of [VGGG16](https://gist.github.com/ksimonyan/211839e770f7b538e2d8#file-readme-md)
or [VGG19](https://gist.github.com/ksimonyan/3785162f95cd2d5fee77#file-readme-md).
2. Open sovlesurgery.py and edit the path of proto and weights
For example: proto = '/home/XYZ/Desktop/DilatedFCNSegmentaiton/VGG_ILSVRC_16_layers_deploy.prototxt'
weights = '/home/XYZ/Dekstop/DilatedFCNSegmentation/VGG_ILSVRC_16_layers.caffemodel'
3. Create a folder inside Dilated_FCN-2s_VGG16 called snapshot. Ex: DilatedFCNSegmentation/Dilated_FCN-2s_VGG16/snapshot
4. Open the solver.protoxt and edit snapshot_prefix. The path will be of the new created model with name.
'/home/XYZ/Desktop/DilatedFCNSegmentation/Dilated_FCN-2s_VGG16/snapshot/vgg16surgery'
5. Now run solvesurgery.py and this will create a file named vgg16surgery_iter_1.caffemodel inside the folder called
Dilated_FCN-2s_VGG16/snapshot