Abstract
A technique for training data augmentation and preprocessing that trains the
network without any specific face detector and makes the trained landmark detector
robust to input face bounding boxes generated by different face detectors during testing.
During testing time, the network can be cascaded with any arbitrary face detector which
can output a reasonable face bounding box and still output good facial landmark
estimation.
Creative Commons License
This work is licensed under a Creative Commons Attribution-Share Alike 4.0 License.
Recommended Citation
INC, HP, "A DATA AUGMENTATION AND PREPROCESSING METHOD FOR ROBUST FACIAL LANDMARK DETECTION WITH DIFFERENT FACE DETECTORS", Technical Disclosure Commons, (November 20, 2018)
https://www.tdcommons.org/dpubs_series/1683