This is a caffe-python implementation on Windows 10 for face alignment.
We implemented two-kind of methods.
Method1 repeat global and local regression after initialization regression
Method2 repeat local refinement regression after initialization regression
| Method | Common | Challenging | Full |
|---|---|---|---|
| Stage(Projection) | 8.24 | 12.56 | 9.07 |
| Stage(Adjustment) | 6.25 | 10.16 | 7.02 |
| Stage(Global1) | 4.66 | 8.20 | 5.35 |
| Stage(Local1) | 3.45 | 6.49 | 4.05 |
| Stage(Global2) | 3.59 | 6.62 | 4.18 |
| Stage(Local2) | 3.29 | 6.14 | 3.85 |
| Stage(Global3) | 3.48 | 6.37 | 4.05 |
| Stage(Local3) | 3.28 | 6.09 | 3.83 |
| Regression(Wild, simple net) | 4.07 | 6.90 | 4.62 |
| Regression(Wild, ResNet50) | 3.72 | 6.44 | 4.25 |
- Clone the repository
git clone https://github.com/hyunsungP/facelignmentregression
- make data files (.h5)
make_wild_input.py
and so on.
-
make data file list
Refer to models/list_train_*.txt -
training
On console window with caffe
caffe train --solver=models/ZF_solver.prototxt --gpu=0
Other network are same.
Change prototxt path in the source code.
test_300w_public.py
Other models will be uploaded.