Skip to content

JamesUnicomb/face_recognition

Repository files navigation

Face Recognition

A face recognition model using the facebook embedder model with dlib. A lasagne MLP is used for distinguishing between faces in the database.

Play the gif Video Link: https://www.youtube.com/watch?v=aa7Rk-uxxjI

Face Messages

For each processed frame a message will be published with a ML predicted face and probability accociated with that matching.

Header header
string[] names
float64[] probability
float64[] distances

Probabilities and distances for each prediction are given for each face measurement for rejection of unknowns.

Adapting to your Database

In the face_recognition/training_data folder, make a list of folders of people you want your algorithm to detect.

For the current code, there are five people: Mark, Jez, Johnson, Sophie and Dobby. Put photos of each person in each of these folders (for your people).

Note: each image can only contain one person.

Once your database is complete, run the script:

python path_to_face_recognition/scripts/train_face_recognition.py

This will save a model in the face_recognition/models folder.

Future Work

Work on a way of recognising unknown faces and rejecting faces not in the database.

About

A face recognition library for ROS.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published