加载OpenCV图像到Scikit Learn

Question

我正在编写一个机器学习脚本，用于拍照并标记。我将数据集放在一个文件夹中，并将它们添加到数组中，创建另一个数组用于标签。当我尝试使用svm.fit时，会出现以下错误：

File "scikit.py", line 43, in <module>
    clf.fit(arrayimg, arraylabel)
  File "/home/mkmeral/.local/lib/python2.7/site-packages/sklearn/svm/base.py", line 151, in fit
    X, y = check_X_y(X, y, dtype=np.float64, order='C', accept_sparse='csr')
  File "/home/mkmeral/.local/lib/python2.7/site-packages/sklearn/utils/validation.py", line 521, in check_X_y
    ensure_min_features, warn_on_dtype, estimator)
  File "/home/mkmeral/.local/lib/python2.7/site-packages/sklearn/utils/validation.py", line 405, in check_array
    % (array.ndim, estimator_name))
ValueError: Found array with dim 3. Estimator expected <= 2.

以下是我编写的脚本：

import cv2
import numpy as py
from sklearn import svm

camera_port = 0
camera = cv2.VideoCapture(camera_port)
ramp_frames = 5

def getImage():
    retval, im = camera.read()
    gray_image = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
    return gray_image

def insertToArray(arrayone, arraytwo, no, true):
    if (true==1):
        directory = "/home/mkmeral/Desktop/opencv/strue/"
        arraytwo.append(1)
    else:
        directory = "/home/mkmeral/Desktop/opencv/sfalse/"
        arraytwo.append(0)

    im = cv2.imread(directory + str(no) + ".png")
    gray_image = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
    arrayone.append(gray_image)

arrayimg = []
arraylabel = []
count = 1
while (count<43):
    insertToArray(arrayimg, arraylabel, count, 1)
    print("True = " , count)
    count = count + 1

count = 0
while (count<43):
    insertToArray(arrayimg, arraylabel, count, 0)
    print("False = ", count)
    count = count + 1

print("Done adding to arrays")
clf = svm.SVC()
print("Done adding to arrayssss")
clf.fit(arrayimg, arraylabel)

print("Done fitting")
for i in xrange(ramp_frames):
    temp = getImage()

testimage = getImage()

clf.predict(testimage)

如何将这些图像适配到Scikit learn中，这会对预测来自网络摄像头拍摄的图像造成问题吗？

- John Doe

Answer 1

我不是图像处理方面的专家，但我猜测你的getImage函数对于每个图像都返回一个二维数组。而sckit-learn则期望每个训练实例都是一个一维数组。假设你的所有图像大小都相同，那么以下方法应该可以解决问题。

def getImage():
    retval, im = camera.read()
    gray_image = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
    return gray_image.flatten()

这将把您的每个图像转换为1d数组。如果您的图像大小不都相同，则需要进行一些图像处理步骤，例如调整大小或下采样。

- piman314