Skip to content

Commit

Permalink
FAQ: weight initialization
Browse files Browse the repository at this point in the history 
warn about the perils of not initializing the weights.
in the future this will be handled by the FCN solving script, but not
just yet.
  • Loading branch information
@shelhamer
shelhamer committed Sep 16, 2016
1 parent b3d1e10 commit a581e48
Showing 1 changed file with 4 additions and 0 deletions.
4 changes: 4 additions & 0 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -71,4 +71,8 @@ Note that in our networks there is only one interpolation kernel per output clas
The alignment is handled automatically by net specification and the crop layer.
It is possible, though less convenient, to calculate the exact offsets necessary and do away with this amount of padding.

**Why are all the outputs/gradients/parameters zero?**: This is almost universally due to not initializing the weights as needed.
To reproduce our FCN training, or train your own FCNs, it is crucial to transplant the weights from the corresponding ILSVRC net such as VGG16.
The included `surgery.transplant()` method can help with this.

**What about FCN-GoogLeNet?**: a reference FCN-GoogLeNet for PASCAL VOC is coming soon.

0 comments on commit a581e48

Please sign in to comment.