【上海校区】python ubuntu lib7 -计算人脸特征向量

python 未结 0 121
品牌整合营销
品牌整合营销 2021-06-22 13:54
悬赏:19
1. 检测人脸

2.提取人脸

3.计算人脸特征向量(用于1:N或1:1比对),使用的是shape predictor

import sys
import os
import dlib
import glob

if len(sys.argv) != 4:
    print(
        "Call this program like this:\n"
        "   ./face_recognition.py shape_predictor_5_face_landmarks.dat dlib_face_recognition_resnet_model_v1.dat ../examples/faces\n"
        "You can download a trained facial shape predictor and recognition model from:\n"
        "    http://dlib.net/files/shape_predictor_5_face_landmarks.dat.bz2\n"
        "    http://dlib.net/files/dlib_face_ ... et_model_v1.dat.bz2")
    exit()

predictor_path = sys.argv[1]
face_rec_model_path = sys.argv[2]
faces_folder_path = sys.argv[3]

# Load all the models we need: a detector to find the faces, a shape predictor
# to find face landmarks so we can precisely localize the face, and finally the
# face recognition model.
detector = dlib.get_frontal_face_detector()
sp = dlib.shape_predictor(predictor_path)
facerec = dlib.face_recognition_model_v1(face_rec_model_path)

win = dlib.image_window()

# Now process all the images
for f in glob.glob(os.path.join(faces_folder_path, "*.jpg")):
    print("Processing file: {}".format(f))
    img = dlib.load_rgb_image(f)

    win.clear_overlay()
    win.set_image(img)

    # Ask the detector to find the bounding boxes of each face. The 1 in the
    # second argument indicates that we should upsample the image 1 time. This
    # will make everything bigger and allow us to detect more faces.
    dets = detector(img, 1)
    print("Number of faces detected: {}".format(len(dets)))

    # Now process each face we found.
    for k, d in enumerate(dets):
        print("Detection {}: Left: {} Top: {} Right: {} Bottom: {}".format(
            k, d.left(), d.top(), d.right(), d.bottom()))
        # Get the landmarks/parts for the face in box d.
        shape = sp(img, d)
        # Draw the face landmarks on the screen so we can see what face is currently being processed.
        win.clear_overlay()
        win.add_overlay(d)
        win.add_overlay(shape)

        # Compute the 128D vector that describes the face in img identified by
        # shape.  In general, if two face descriptor vectors have a Euclidean
        # distance between them less than 0.6 then they are from the same
        # person, otherwise they are from different people. Here we just print
        # the vector to the screen.
        face_descriptor = facerec.compute_face_descriptor(img, shape)
        print(face_descriptor)
        # It should also be noted that you can also call this function like this:
        #  face_descriptor = facerec.compute_face_descriptor(img, shape, 100)
        # The version of the call without the 100 gets 99.13% accuracy on LFW
        # while the version with 100 gets 99.38%.  However, the 100 makes the
        # call 100x slower to execute, so choose whatever version you like.  To
        # explain a little, the 3rd argument tells the code how many times to
        # jitter/resample the image.  When you set it to 100 it executes the
        # face descriptor extraction 100 times on slightly modified versions of
        # the face and returns the average result.  You could also pick a more
        # middle value, such as 10, which is only 10x slower but still gets an
        # LFW accuracy of 99.3%.


        dlib.hit_enter_to_continue()

---------------------
作者:_iorilan
来源:CSDN
原文:https://blog.csdn.net/lan_liang/article/details/84715272
版权声明:本文为博主原创文章,转载请附上博文链接!

相关标签:
回答
  • 消灭零回复
提交回复