Loading presentation...

Present Remotely

Send the link below via email or IM


Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.


Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

DeepFace Architecture

No description

Marwa Said

on 4 July 2014

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of DeepFace Architecture

DeepFace Architecture
3 D Face Modeling
Pose Estimation From Still Images
1- 3 D Face Scans
350 3D scans
USF Human Database
2- Registration
To establish Correspondence
Algorithm Used: ICP (Iterative Closest Point)
3- Model Building
Principale Component Analyses PCA
4- Model Fitting
Open Source MATLAB code- Dec. 2013
The Facebook AI Guy: Yaniv Taigman
Pose Estimation and Face Alignment
Step 1: 2D Alignment
1- Face Detection:
Using Viola- Jones Detector
2- Localization of 6 Facial Landmarks:
Center of the eyes, tip of the nose, and mouth locations (Xiangxin Zhu and Deva Ramanan)
3- Rotate and translate to the 6 anchor locations:
Similarity (Euclidean) Transformations
2D alignment Result
Step 1: 2D Alignment
Step 2: 3D Alignment
1- "67" Landmarks Localization
Step 2: 3D Alignment
1- Landmarks Localization
67 landmarks is localized in the 2D aligned corp using the method by (Xiangxin Zhu and Deva Ramanan)
Step 2: 3D Alignment
2- Pre-processing of a 3D Generic Model
Step 2:3D Alignment
2- Pre- Processing of a 3D Generic Model
- Take the average of the 3D scans from The USF database,
- Post processed to be aligned as same as the query image Using Camera Motion Modeling and calibration (Multiple View Geometry),
Step 2: 3D Alignment
3- the"67" landmarks Localization at the 3D generic shape
Step 2: 3D Alignment
3- the"67" landmarks Localization at the 3D generic shape
Manually Place the 67 facial landmarks on the 3D generic shape to achieve full correspondence between the 2 shapes (Using the same method as before).
Step 2: 3D Alignment
4- 3D Pose Estimation (Frontization) and 2D projection
4- 3D Pose Estimation (Frontization) and 2D projection
Step 2: 3D Alignment
- Pose Estimation and frontization through Camera Motion modeling and calibration (Multiple View Geometry)
- 2D projection (Affine transformation and projective geometry)
Final Result
How does this relate to the MATLAB code we have got??
Viewing Real World Faces in 3D
Landmarks Localization
The Use of a 3D Generic Model
Camera Calibration and Pose Estimation
Full transcript