1 / 18

Extensive Facial Landmark Localization with Coarse-to-fine Convolutional Neural Network

Extensive Facial Landmark Localization with Coarse-to-fine Convolutional Neural Network. Erjin Zhou Haoqiang Fan Zhimin Cao Yuning Jiang Qi Yin Megvii Inc., Beijing. Problem. 2-Eyes Detection. Problem. 5-Corners Detection. 2-Eyes Detection. Problem. 23-Points Detection. Problem.

laith-grant
Download Presentation

Extensive Facial Landmark Localization with Coarse-to-fine Convolutional Neural Network

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Extensive Facial Landmark Localization with Coarse-to-fine Convolutional Neural Network Erjin Zhou Haoqiang Fan Zhimin Cao Yuning Jiang Qi Yin Megvii Inc., Beijing

  2. Problem 2-Eyes Detection

  3. Problem 5-Corners Detection 2-Eyes Detection

  4. Problem 23-Points Detection

  5. Problem More and more landmarks are required! 68-Points Detection

  6. Idea I Single Predictor Disadvantage: the difficulty of localizing each point is quite different, and it is hard to optimize all points by a single model.

  7. Idea II Predictor 1 Predictor 2 Predictor 3 Predictor 4 Predictor 5 …… Disadvantages: no geometric constrains; heavy computational burden.

  8. Idea III Predictor 1 Predictor 2 Predictor 3 Predictor 4 … Advantage: component contexts are considered.

  9. Observation However, do we really need the mouth to locate the eye corners?

  10. Our Intuition Predictor 1 Predictor 2 Component estimator Predictor 3 Predictor 4 …

  11. Framework

  12. Experiments

  13. Experiments

  14. Experiments Level 2 Level 3 Level 4 Average error on each level of CNN.

  15. Results on 300-W

  16. Results on 300-W

  17. About us

  18. Thanks!

More Related