A simple face aligment method, based on pytorch
The keypoint model encodes and decodes the x and y coordinates using heatmap and offset of x and y, achieving SOTA on WFLW dataset. Like object detection, heatmap predicts which point is a positive sample on the featuremap, represented as a highlighted area, while x and y offsets are responsible for predicting the specific coordinates of these positive samples. And it achieves ** NME 3.95 on WFLW ** with no extern data. Contact me if u have problem about it. [email protected] :)
demo pictures:
this gif is from github.com/610265158/Peppa_Pig_Face_Engine )
WLFW test set.
| WFLW | inputsize | NME | Flops(G) | Params(M) | Pose | Exp. | Ill. | Mu. | Occ. | Blur | pretrained |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Student | 128x128 | 4.80 | 0.35 | 3.25 | 8.53 | 5.00 | 4.61 | 4.81 | 5.80 | 5.36 | skps |
| Teacher | 128x128 | 4.17 | 1.38 | 11.53 | 7.14 | 4.32 | 4.01 | 4.03 | 4.98 | 4.68 | skps |
| Student | 256x256 | 4.35 | 1.39 | 3.25 | 7.53 | 4.52 | 4.16 | 4.21 | 5.34 | 4.93 | skps |
| Teacher | 256x256 | 3.95 | 5.53 | 11.53 | 7.00 | 4.00 | 3.81 | 3.78 | 4.85 | 4.54 | skps |
-
pytorch
-
opencv
-
timm
-
Download WFLW data. Set them in train_config.py.
-
then
run.sh -
by default it is trained with mobilenetv3-large as student, efficientnetb5 as teacher.
python tools/eval_WFLW.py --weight xxx.pth --data_dir ./ --img_size 256
python vis.py --model ./keypoints.pth
python vis.py --model ./keypoints.pth
python tools/convert_to_onnx.py --model ./keypoints.pth