Convenient API of MTCNN face or face landmark detection.

Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Neural Networks

Tensorflow implementation of the face detection / alignment algorithm found at https://github.com/kpzhang93/MTCNN_face_detection_alignment

I will going to support more framework such as Caffe, Pytorch and Mxnet.

• Tensorflow
• Caffe
• Pytorch
• Mxnet

Requirement

• tensorflow
• python2.x or python3.x
• cv2

This is a python package and you should import MTCNN_API.

MTCNN_API.detect_cv2_ims(images,minsize=20,threshold=(0.6, 0.7, 0.7),scale_factor=0.709,gpu_fraction=0.5)

• images: a list contains cv2 opened images
• minsize: minimum size of face
• threshold: three steps's threshold
• scale_factor: scale factor
• gpu_fraction: tensorflow gpu_fraction

return: (boxes, landmarks)

• boxes is a list contains ndarray with shape (n_faces_in_pic, 5), 5 number represent for (x1,y1,x2,y2,score).
• landmarks is a list contains ndarray with shape(n_faces_in_pic, 10),10 landmark number represent for (leyex,reyex,nosex,lmouthx,rmouthx,leyey,reyey,nosey,lmouthy,rmouthy) l:Left, r:Right

import cv2
import MTCNN_API

images = []
for file_name in ['test1.jpg', 'test2.png', 'test3.png']:
    im = cv2.imread(file_name)
    images.append(im)

boxes, landmarks = MTCNN_API.detect_cv2_ims(images)