从图像中精确地雕刻出人类答案

【问题标题】：Carving Out exactly humans From Image从图像中精确地雕刻出人类
【发布时间】：2021-04-23 11:45:42
【问题描述】：

我想从目前能够在人类周围创建一个盒子的图像中切出精确的人类我使用的代码如下：

import cv2
import numpy as np
import imutils

protopath = r"C:\Users\Admin\MobileNetSSD_deploy.prototxt"
modelpath = r"C:\Users\Admin\MobileNetSSD_deploy.caffemodel"
detector = cv2.dnn.readNetFromCaffe(prototxt=protopath, caffeModel=modelpath)

CLASSES = ["background", "aeroplane", "bicycle", "bird", "boat",
           "bottle", "bus", "car", "cat", "chair", "cow", "diningtable",
           "dog", "horse", "motorbike", "person", "pottedplant", "sheep",
           "sofa", "train", "tvmonitor"]


def main():
    image = cv2.imread(r'E:\trial2.jpg')
    image = imutils.resize(image,300)

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

    blob = cv2.dnn.blobFromImage(image, 0.007843, (W, H), 127.5)

    detector.setInput(blob)
    person_detections = detector.forward()

    for i in np.arange(0, person_detections.shape[2]):
        confidence = person_detections[0, 0, i, 2]
        if confidence > 0.5:
            idx = int(person_detections[0, 0, i, 1])

            if CLASSES[idx] != "person":
                continue
                

            person_box = person_detections[0, 0, i, 3:7] * np.array([W, H, W, H])
            (startX, startY, endX, endY) = person_box.astype("int")
            cv2.rectangle(image, (startX, startY), (endX, endY), (0, 0, 255), 2)

    cv2.imshow("Results", image)
    cv2.waitKey(0)
    cv2.destroyAllWindows()
main()

我想要什么 我想准确地勾勒出人类的轮廓

任何帮助或提示将不胜感激

编辑 1：由 Sunil Kumar 建议

import numpy as np
import cv2
from matplotlib import pyplot as plt

img = cv2.imread(r'E:\trial3.jpg')
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
ret, thresh = cv2.threshold(gray,0,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
# noise removal
kernel = np.ones((3,3),np.uint8)
opening = cv2.morphologyEx(thresh,cv2.MORPH_OPEN,kernel, iterations = 2)

# sure background area
sure_bg = cv2.dilate(opening,kernel,iterations=3)

# Finding sure foreground area
dist_transform = cv2.distanceTransform(opening,cv2.DIST_L2,5)
ret, sure_fg = cv2.threshold(dist_transform,0.7*dist_transform.max(),255,0)

# Finding unknown region
sure_fg = np.uint8(sure_fg)
unknown = cv2.subtract(sure_bg,sure_fg)
# Marker labelling
ret, markers = cv2.connectedComponents(sure_fg)

# Add one to all labels so that sure background is not 0, but 1
markers = markers+1

# Now, mark the region of unknown with zero
markers[unknown==255] = 0
markers = cv2.watershed(img,markers)
img[markers == -1] = [255,0,0]
cv2.imshow("Results", img)
cv2.waitKey(0)
cv2.destroyAllWindows()

以及我得到的当前输出

【问题讨论】：

SSD网络是一个物体检测模型； “对象检测”通常被定义为边界框。您要求的是“掩码”，通常这个问题称为“实例分割”，您需要运行不同的模型，如 Mask R-CNN，而不是 YOLO 或 SSD 等。请参阅此处了解基本信息概述：towardsdatascience.com/…
@isAif 那是我目前的结果......预期的结果是在人类身上创建轮廓

标签： python-3.x image-segmentation opencv-python

【解决方案1】：

你要找的是instance segmentation，比物体检测准确多了：
而对象检测仅在每个人周围输出一个边界框（正如您已经注意到的）。实例分割尝试为每个实例输出一个精确的掩码 - 这就是您要查找的内容。

我会尝试以下实例分割框架：

【讨论】：