Face Similarity searching ~ landmark detecting

Thanks to this post of facial landmarks and the openface project!

11/11  updated the image pool to 710000

11/3  updated the image pool to 540000

10/14 add face similarity searching! from a 4000-photo pool

Deep Learning model find 128 features of each face –Then Cosine distance ~ simple but powerful. 

face similarity searching from celebrities (which superstar looks like you the most)?

I’ve implemented the idea to a web app that can batch processing face chopping from pics  & facial landmarks pinpointing!

The next step will be training beauty score rating model & a face similarity comparing to superstars (which star looks like you the most)?

Original App Url

Machine Learning: Regression with GraphLab Create

bias-variance trade-off gradient descent
Ridge regression        cross-validation      measure of fit + measure of  model complexity  
Lasso regression coordinate descent feature selection measure of fit + (different) measure of model complexity
Nearest Neighbor Regression & Kernel Regression
 
concave  hax Max value
convex  has Min value
Models

 

• Linear regression
• Regularization: Ridge (L2), Lasso (L1)
• Nearest neighbor and kernel regression
Algorithms • Gradient descent
• Coordinate descent
Concepts

 

• Loss functions, bias-variance tradeoff,

cross-validation, sparsity, overfitting,

model selection, feature selection