Skip to content
Face identification with CNN + TPE using Keras
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
utils
.gitignore
LICENSE
README.md
bottleneck.py
cnn.py
demo_app0.py
demo_app1.py
identification.py
model.py
preprocessing.py
test_cnn.py
test_tpe.py
tpe.py
train_cnn.py
train_tpe.py

README.md

Face identification using CNN + TPE demo

Face identification demo which implements the Triplet Probabilistic Embedding for Face Verification and Clustering work.

demo app screenshot

Requirements

  • python3
  • keras
  • dlib
  • scikit-image

Usage

Get all the files from here and put them to the model dir. Then start demo_app1.py.

Training

Download the face_template.npy and shape_predictor_68_face_landmarks.dat from here and put them to the model dir.

Place training data in following order:

data\
    dev_protocol.npy
    dev\
        1.jpg
        2.jpg
        3.jpg
        ...
    test\
        subject_0\
            1.jpg
            2.jpg
            ...
        subject_1\
            1.jpg
            2.jpg
            ...
        ...
    train\
        subject_0\
            1.jpg
            2.jpg
            ...
        subject_1\
            1.jpg
            2.jpg
            ...
        ...

Then run as follows:

  1. utils/load_data.py
  2. train_cnn.py
  3. train_tpe.py

Use the test scripts to test your model.

You can’t perform that action at this time.