裁剪图像后转换 Numpy 数组

Question

我目前正在使用此代码：

def bundle_contour(image):
    src_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    crop_img = cv2.resize(src_gray, (196, 196))
    v = np.median(crop_img)
    lower = int(max(0, (1.0 - 0.33) * v))
    upper = int(min(255, (1.0 + 0.33) * v))
    filter = cv2.bilateralFilter(src_gray, 9, 75, 75)
    filter = filter.astype(np.uint8)
    edged = cv2.Canny(filter, lower, upper)
    kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(3,3))
    thresh = cv2.dilate(edged, kernel, iterations=2)

    # Find contours in threshold image, then grab the largest one
    cnts = cv2.findContours(thresh.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    cnts = imutils.grab_contours(cnts)
    c = max(cnts, key=cv2.contourArea)

    # Find the extreme points
    extLeft = tuple(c[c[:, :, 0].argmin()][0])
    extRight = tuple(c[c[:, :, 0].argmax()][0])
    extTop = tuple(c[c[:, :, 1].argmin()][0])
    extBot = tuple(c[c[:, :, 1].argmax()][0])

    # crop new image out of the original image using the four extreme points (left, right, top, bottom)
    new_image = image[extTop[1]:extBot[1], extLeft[0]:extRight[0]]
    return new_image

def load_data(filename):
    image = cv2.imread(filename)
    # crop the bundle and ignore the unnecessary rest part of the image
    cnt_image = bundle_contour(image)
    # resize image
    resized_image = cv2.resize(cnt_image, dsize=(196, 196), interpolation=cv2.INTER_CUBIC)
    # normalize values
    norm_image = resized_image / 255.
    reshaped_img = norm_image.reshape(196,196,1)
    return reshaped_img

def compose_dataset(df):
    data = []
    labels = []

    for img_path, label in df.values:
        data.append(load_data(img_path))
        labels.append(label)

    return np.array(data), np.array(labels)

X_train, y_train = compose_dataset(train_df)
print('Train data shape: {}, Labels shape: {}'.format(X_train.shape, y_train.shape))

我收到此错误：

ValueError：无法将大小为 115248 的数组重塑为形状 (196,196,1)。

如果我将代码更改为norm_image.reshape(196,196,-1) ，我的 output 是：

Train data shape: (134, 196, 196, 3), Labels shape: (134,)

我怎样才能使输入到我的 CNN 的形状为 (196,196,1)？

Answer 1

@HansHirse 完全正确。 我正在写一个答案以使其更具体。

该错误很可能是由load_data方法引起的。 由于您正在读取图像但未转换为灰度。

因此，您需要添加句子

image = cv2.cvtColor(image, cv2.BGR2GRAY)

到load_data 。 所以正确的代码是：

def load_data(filename):
    image = cv2.imread(filename)
    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    # crop the bundle and ignore the unnecessary rest part of the image
    cnt_image = bundle_contour(image)
    # resize image
    resized_image = cv2.resize(cnt_image, dsize=(196, 196), interpolation=cv2.INTER_CUBIC)
    # normalize values
    norm_image = resized_image / 255.
    reshaped_img = norm_image.reshape(196,196,1)
    return reshaped_img

如果您在 load_data 之后使用bundle_contour load_data ，则不需要这句话：

src_gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

你可以删除它。