/face_recognition_with_siamese_network

A face recognition model is trained by using Transfer Learning and applying Siamese Network and Triplet Loss principles.

Primary LanguageJupyter Notebook

A Face Reognition model that is developed by using:

  • Transfer Learning. ResNet50 model (using weights of imagenet from Keras) is used in the project.
  • principles of Siamese Network and Triplet Loss methods.
  • Keras and Tensorflow

The model is trained based on face images from 8 different peoples. The dataset that is used in training is not shared in order not to violate any rights. But the model can still be used/trained with any face image dataset.

The theory and implementation aspects of the project have been described in great detail here: Theory: http://www.loheden.com/2018/07/face-recognition-with-siamese-network.html Implementation aspects: http://www.loheden.com/2018/07/face-recognition-with-siamese-network_29.html

The project has just been developed for educational purposes, it addresses almost all challenges that one can face during the development of such a project. Note that the code is not fully optimized.