The project is a part of Udacity Computer Vision Nanodegree. It contains the algorithms for detecting 68 specific facial keypoints. The system comprises from
- Haar Cascades for face detection,
- CNN Resnet18 for grayscale images for facial keypoints prediction.
Pretrained Resnet18 was additionally trained on set of 3462 images
Training images are shown in grayscale mode, with green keypoints.
Test images are shown in grayscale mode, with green keypoints, and red predicted keypoints.