face_clustering

This project aims at clustering people by face. Author: Paulo Renato Conceição Mendes

Install:

You can use pip to install.

pip install https://github.com/TeleMidia/face_clustering/releases/download/0.0.1/face_clustering-0.0.1-py3-none-any.whl

Usage:

from faces_clustering import Clusterer

clusters = Clusterer(n_clusters=4, urls=complete_urls).clusterize()

Basic Params

• n_clusters: number of clusters used

• urls: list of urls from the images you want to clusterize

Additional Params

• backbone: string that specifies the backbone used for extracting face features. It can be one of the following:

'senet50'(default), 'resnet50', 'vgg16'

clusters = Clusterer(n_clusters=4, urls=complete_urls, backbone='resnet50').clusterize()

• algs: list that specify which clustering algorithms you want to use. By default, all of them are used:

['kmeans', 'gmm', 'affinity', 'agglomerative']

clusters = Clusterer(n_clusters=4, urls=complete_urls, algs=['kmeans', 'gmm']).clusterize()

• kmeans refers to sklearn kmeans
• gmm refers to sklearn Gaussian Mixture Model
• affinity refers to sklearn Affinity Propagation
• agglomerative refers to sklearn Ward Agglomerarive

References

• Keras-vggface used for feature extraction: https://github.com/rcmalli/keras-vggface
• MTCNN for face detection: https://pypi.org/project/mtcnn/
• Sklearn for clustering algorithms: https://scikit-learn.org/stable/modules/clustering.html#clustering