no code implementations • CVPR 2020 • Bjorn Browatzki, Christian Wallraven
Current supervised methods for facial landmark detection require a large amount of training data and may suffer from overfitting to specific datasets due to the massive number of parameters.