Opencv Random Cropping

准备机器学习的训练数据第二步——随机裁剪(Random Cropping)

下面,通过从imori1.jpg中随机裁剪图像制作训练数据。

这里,从图像中随机切出200个60\times60的矩形。

并且,满足下面的条件:

  1. 使用np.random.seed(0),求出裁剪的矩形的左上角座标x1 = np.random.randint(W-60)y1=np.random.randint(H-60)
  2. 如果和 Ground-truth (gt = np.array((47, 41, 129, 103), dtype=np.float32))的\text{IoU}大于0.5,那么就打上标注1,小于0.5就打上标注0

答案中,标注1的矩形用红色画出,标注0的矩形用蓝色的线画出,Ground-truth用绿色的线画出。我们简单地准备蝾螈头部和不是头部的图像。

输入 (imori_1.jpg) 输出

python实现:

import cv2
import numpy as np

np.random.seed(0)

# get IoU overlap ratio
def iou(a, b):
    # get area of a
    area_a = (a[2] - a[0]) * (a[3] - a[1])
    # get area of b
    area_b = (b[2] - b[0]) * (b[3] - b[1])

    # get left top x of IoU
    iou_x1 = np.maximum(a[0], b[0])
    # get left top y of IoU
    iou_y1 = np.maximum(a[1], b[1])
    # get right bottom of IoU
    iou_x2 = np.minimum(a[2], b[2])
    # get right bottom of IoU
    iou_y2 = np.minimum(a[3], b[3])

    # get width of IoU
    iou_w = iou_x2 - iou_x1
    # get height of IoU
    iou_h = iou_y2 - iou_y1

    # get area of IoU
    area_iou = iou_w * iou_h
    # get overlap ratio between IoU and all area
    iou = area_iou / (area_a + area_b - area_iou)

    return iou


# crop and create database
def crop_bbox(img, gt, Crop_N=200, L=60, th=0.5):
    # get shape
    H, W, C = img.shape

    # each crop
    for i in range(Crop_N):
        # get left top x of crop bounding box
        x1 = np.random.randint(W - L)
        # get left top y of crop bounding box
        y1 = np.random.randint(H - L)
        # get right bottom x of crop bounding box
        x2 = x1 + L
        # get right bottom y of crop bounding box
        y2 = y1 + L

        # crop bounding box
        crop = np.array((x1, y1, x2, y2))

        # get IoU between crop box and gt
        _iou = iou(gt, crop)

        # assign label
        if _iou >= th:
            cv2.rectangle(img, (x1, y1), (x2, y2), (0,0,255), 1)
            label = 1
        else:
            cv2.rectangle(img, (x1, y1), (x2, y2), (255,0,0), 1)
            label = 0

    return img

# read image
img = cv2.imread("imori_1.jpg")

# gt bounding box
gt = np.array((47, 41, 129, 103), dtype=np.float32)

# get crop bounding box
img = crop_bbox(img, gt)

# draw gt
cv2.rectangle(img, (gt[0], gt[1]), (gt[2], gt[3]), (0,255,0), 1)

cv2.imwrite("out.jpg", img)
cv2.imshow("result", img)
cv2.waitKey(0)


赞(0)

评论 抢沙发

  • 昵称 (必填)
  • 邮箱 (必填)
  • 网址

Opencv实例

最新文章