Detect, recognize and verify faces using hybrid features: “deep” features from VGG-net + HoG + LBP. Uses VGG Face matconvnet for deep feature extraction along with features obtained from traditional methods HoG and LBP. Hybrid Features help increase accuracy of recognition to a significant level.
Can detect most faces. (Warning may have bias towards the data we trained it with)
Can recognize the following faces:
- Abdullah Gul
- Amelie Mauresmo
- Andy Roddick
- Angelina Jolie
- Atal Bihari
- Bill Gates
- Bill Simon
- Britney Spears
- Carlos Menem
- Carlos Moya
- David Nalbandian
- Dick Cheney
- Dominique de
- Eduardo Duhalde
- Fidel Castro
- George Robertson
- Halle Berry
- Hamid Karzai
- Hillary Clinton
- Hu Jintao
- Igor Ivanov
- James Blake
- Jean Charest
- Jennifer Aniston
- Jennifer Lopez
- Jeremy Greenstock
- Jiang Zemin
- John Bolton
- John Howard
- John Kerry
- John Snow
- Joschka Fischer
- Jose Maria
- Julianne Moore
- Julie Gerberding
Trained with approximately 10-15 cropped face images of size 64x64 for each person.
Can verify if it is the same person given two images.
Trained with 1800 records each containing two faces and ground truth value of whether they were same or not.
Yes...! All that's required is a few images of your face. Train the model with the same and Voila! More images = More Accurate verification
This mini-project was submitted for Image and Vision Computing Assignment
Assignment_1.m:
Part 1: Face Detection
Assignment_2.m:
Part 2: Face Recognition and Verification
bonus.m:
Bonus marks: Linked detector and recogniser, indentifies multiple individuals (drawn from validation dataset which the model has not seen before) correctly.
We split Assignment_2.m into two files for developmental convenience
Assignment_2_Part1.m:
Face Recognition
Assignment_2_Part2.m:
Face Verification
VGG Face model Location: ./library/matconvnet/data/models/vgg-face.mat
We made separate files for demo which loads the trained models
Assignment_1_eval.m:
Face Detection
Loads model from ./face_detector.mat
Assignment_2_Part1_eval.m:
Face Recognition
Loads model from ./models/fr_model.mat
Assignment_2_Part2_eval.m:
Face Verification
Loads models from ./models/fv_model.mat