《计算机视觉40列》K均值聚类20、24

20.2-20.3

import numpy as np
import cv2
import matplotlib.pyplot as plt
x = np.random.randint(0,100,(50,2))
x = np.float32(x)

criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 10, 1.0)
ret, label, center = cv2.kmeans(x, 2, None, criteria, 10, cv2.KMEANS_RANDOM_CENTERS)

print("距离：", ret)
print("标签：", np.reshape(label, -1))
print("分类中心点：\n", center)

A = x[label.ravel()==0]
B = x[label.ravel()==1]

plt.scatter(A[:,0], A[:,1], c = 'g', marker = 's')
plt.scatter(B[:,0], B[:,1], c = 'r', marker = 'o')

plt.scatter(center[0,0], center[0,1],s = 200,c = 'b', marker = 's')
plt.scatter(center[1,0], center[1,1],s = 200,c = 'b', marker = 'o')
plt.show()

import numpy as np

import cv2



img = cv2.imread('cat.jpg')

data = img.reshape((-1,3))

data = np.float32(data)



criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 10, 1.0)

K = 2

ret, label, center = cv2.kmeans(data, K, None, criteria, 10, cv2.KMEANS_RANDOM_CENTERS)



print("距离：", ret)

print("标签：\n", label)

print("分类中心点：\n", center)



center = np.uint8(center)

res1 = center[label.flatten()]

res2 = res1.reshape((img.shape))

cv2.imshow("original", img)

cv2.imshow("result", res2)

cv2.waitKey()

cv2.destroyAllWindows()

列24.1

import numpy as np

import cv2


image = cv2.imread("tower.jpg")


config = "model/bvlc_googlenet.prototxt"

model = "model/bvlc_googlenet.caffemodel"

net = cv2.dnn.readNetFromCaffe(config, model)

blob = cv2.dnn.blobFromImage(image, 1, (244, 244), (104, 117, 123))

net.setInput(blob)

prob = net.forward()


classes = open("model/label.txt", "rt").read().strip().split("\n")

rowIndex = np.argsort(prob[0])[::-1][0]



result = "result:{},{:.0f}%".format(classes[rowIndex],prob[0][rowIndex]*100)

cv2.putText(image, result, (25, 45), cv2.FONT_HERSHEY_SIMPLEX,1, (0, 0, 255), 2)



cv2.imshow("result", image)

cv2.waitKey()

cv2.destroyAllWindows()

24.2-24.3

import numpy as np

import cv2


image = cv2.imread("test2.jpg")


classes = open('coco.names', 'rt').read().strip().split("\n")


net = cv2.dnn.readNetFromDarknet("yolov3.cfg","yolov3.weights")

blob = cv2.dnn.blobFromImage(image, 1.0 / 255.0, (416, 416), (0, 0, 0), True, crop=False)


net.setInput(blob)


outInfo = net.getUnconnectedOutLayersNames()

outs = net.forward(outInfo)


resultIDS = []

boxes = []

confindences = []

(H, W) = image.shape[:2]

for out in outs:

    for candidate in out:

        scores = candidate[5:]

        classID = np.argmax(scores)

        confindence = scores[classID]



        if confindence > 0.5:



            box = candidate[0:4] * np.array([W, H, W, H])



            (centerX, centerY, width, height) = box.astype("int")



            x = int(centerX - (width / 2))

            y = int(centerX - (height / 2))



            boxes.append([x, y, int(width), int(height)])

            confindences.append(float(confindence))

            resultIDS.append(classID)



indexes = cv2.dnn.NMSBoxes(boxes, confindences, 0.5, 0.4)

classesCOLORS = np.random.randint(0, 255, size = (len(classes), 3), dtype="uint8")

for i in range(len(boxes)):

    if i in indexes:

        x, y, w, h = boxes[i]

        color = [int(c) for c in classesCOLORS[resultIDS[i]]]

        cv2.rectangle(image, (x, y), (x+w, y+h), color, 2)

        result = "{}: {:.0f}%".format(classes[resultIDS[i]], confindences[i]*100)

        cv2.putText(image, result, (x, y+35), cv2.FONT_HERSHEY_SIMPLEX, 1, color, 2)



cv2.imshow('result', image)

cv2.waitKey(0)

cv2.destroyAllWindows()

import numpy as np
import cv2

image = cv2.imread("test2.jpg")
(H, W) = image.shape[:2]
classes = open('object_detection_classes_pascal_voc.txt', 'rt').read().strip().split("\n")

classesCOLORS = (np.random.uniform(0, 255, size=(len(classes), 3)))

config = "MobileNetSSD_deploy.prototxt.txt"
model = "MobileNetSSD_deploy.caffemodel"
net = cv2.dnn.readNetFromCaffe(config, model)
blob = cv2.dnn.blobFromImage(cv2.resize(image, (300, 300)), 0.007843, (300, 300), 127.5)

net.setInput(blob)
outs = net.forward()
print(outs.shape)

import cv2
import numpy as np

image = cv2.imread("a.jpg")
H, W = image.shape[:2]

classes = open("object_detection_classes_pascal_voc.txt",
               "rt").read().strip().split("\n")

classesCOLORS = np.random.randint(0, 255, size=(len(classes), 3), dtype = "uint8")

classesCOLORS[0] = (0,0,0)

rowHeight = 30

colorChart = np.zeros((rowHeight * len(classesCOLORS), 200, 3), np.uint8)

for i in range(len(classes)):
    row = colorChart[i * rowHeight:(i + 1) * rowHeight]
    row[:,:] = classesCOLORS[i]
    cv2.putText(row, classes[i], (0, rowHeight//2),
                cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255))

cv2.imshow('colorChart', colorChart)

model = "fcn8s-heavy-pascal.caffemodel"
config = "fcn8s-heavy-pascal.prototxt"
net = cv2.dnn.readNet(model, config)
blob = cv2.dnn.blobFromImage(image, 1.0, (W, H), (0, 0, 0), False, crop=False)
net.setInput(blob)
score = net.forward()
classIDS = np.argmax(score[0], axis=0)
print(classIDS.shape)

mask = np.stack([classesCOLORS[index] for index in classIDS.flatten()])
mask = mask.reshape(H, W, 3)

result = cv2.addWeighted(image, 0.2, mask, 0.8, 0)
cv2.imshow("result", result)
cv2.waitKey(0)
cv2.destroyWindow()

import cv2
import numpy as np

image = cv2.imread('e.jpg')

(H, W) = image.shape[:2]
background = np.zeros((H, W, 3), np.uint8)
background[:] = (100, 100, 0)

LABELS = open('object_detection_classes_coco.txt').read().strip().split('\n')

net = cv2.dnn.readNetFromTensorflow('dnn/frozen_inference_graph.pb',
                                    'dnn/mask_rcnn_inception_v2_coco_2018_01_28.pbtxt')
blob = cv2.dnn.blobFromImage(image, swapRB=True)
net.setInput(blob)
boxes, masks = net.forward(['detection_out_final', 'detection_masks'])

number = boxes.shape[2]

for i in range(number):
    classID = int(boxes[0, 0, i,1])
    confidence = boxes[0, 0, i, 2]
    if confidence > 0.5:
        box = boxes[0, 0, i, 3:7] * np.array([W, H, W, H])
        (x1, y1, x2, y2) = box.astype('int')
        box = background[y1: y2, x1: x2]
        boxHeight = y2 - y1
        boxWidth = x2 - x1
        mask = masks[i, int(classID)]
        mask = cv2.resize(mask, (boxWidth, boxHeight))
        rst, mask = cv2.threshold(mask, 0.5, 255, cv2.THRESH_BINARY)
        contours, hierarchy = cv2.findContours(np.array(mask, np.uint8),
                                                cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
        color = np.random.randint(0, 255, 3)
        color = tuple([int(x) for x in color])
        cv2.drawContours(box, contours, -1, color, -1)
        msg = '{}:{:.0f} %'.format(LABELS[classID], confidence * 100)
        cv2.putText(background, msg, (x1 + 50, y1 + 45),
                    cv2.FONT_HERSHEY_SIMPLEX, 0.8, (255, 255, 255), 2)

result = cv2.addWeighted(image, 0.2, background, 0.8, 0)
cv2.imshow('original', image)
cv2.imshow('result', result)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2



image = cv2.imread("tute.jpg")

H, W = image.shape[:2]



net = cv2.dnn.readNetFromTorch('model/eccv16/starry_night.t7')

blob = cv2.dnn.blobFromImage(image, 1.0, (W, H), (0, 0, 0), swapRB=False, crop=False)



net.setInput(blob)

out = net.forward()

print(out)



out = out.reshape(out.shape[1], out.shape[2], out.shape[3])



cv2.normalize(out, out, alpha=0.0, beta=1.0, norm_type=cv2.NORM_MINMAX)



result = out.transpose(1, 2, 0)



cv2.imshow("original", image)

cv2.imshow('result', result)

cv2.waitKey()

cv2.destroyWindow()