YOLO区域目标检测

基于YOLO的区域检测

最近在研究YOLO的项目，也研究一些大神关于区域检测的代码，发现比较稀缺，几番周折之后，研究了一下，搞了一套暂且能凑合用的，先码上，免得以后自己忘了。

---

一、使用步骤

1.判断点是否在多边形内部的

代码如下：

def is_poi_in_poly(pt, poly):
    """
    判断点是否在多边形内部的 pnpoly 算法
    :param pt: 点坐标 [x,y]
    :param poly: 点多边形坐标 [[x1,y1],[x2,y2],...]
    :return: 点是否在多边形之内
    """
    nvert = len(poly)
    vertx = []
    verty = []
    testx = pt[0]
    testy = pt[1]
    for item in poly:
        vertx.append(item[0])
        verty.append(item[1])
    j = nvert - 1
    res = False
    for i in range(nvert):
        if (verty[j] - verty[i]) == 0:
            j = i
            continue
        x = (vertx[j] - vertx[i]) * (testy - verty[i]) / (verty[j] - verty[i]) + vertx[i]
        if ((verty[i] > testy) != (verty[j] > testy)) and (testx < x):
            res = not res
        j = i
    return res

2.检测物体是否在多边形危险区域内

代码如下：

def in_poly_area_dangerous(xyxy,area_poly):
    """
    检测人体是否在多边形危险区域内
    :param xyxy: 人体框的坐标
    :param img_name: 检测的图片标号，用这个来对应图片的危险区域信息
    :return: True -> 在危险区域内，False -> 不在危险区域内
    """
    # print(area_poly)
    if not area_poly:  # 为空
        return False
    # 求物体框的中点
    object_x1 = int(xyxy[0])
    object_y1 = int(xyxy[1])
    object_x2 = int(xyxy[2])
    object_y2 = int(xyxy[3])
    object_w = object_x2 - object_x1
    object_h = object_y2 - object_y1
    object_cx = object_x1 + (object_w / 2)
    object_cy = object_y1 + (object_h / 2)
    return is_poi_in_poly([object_cx, object_cy], area_poly)

---

3.推理中的修改

代码如下：

  if det is not None and len(det):
        # Rescale boxes from img_size to im0 size
        det[:, :4] = scale_coords(img.shape[2:], det[:, :4], image.shape).round()
        for *xyxy, conf, cls in det:
            image_result = {}
            if names[int(cls)] == "black_car":
                if in_poly_area_dangerous(xyxy,area_poly) == True:

---

4.调用

代码如下：

if __name__ == '__main__':
    area_poly = [[6,138],[1277,133],[1275,718],[4,719]]   #坐标点
    ModelPath = './model/best.pt'
    Device = select_device('cpu')  # # cpu:cpu,  GPU:0,1,2
    Model = load_model(ModelPath, Device)
    ImgPath = '71.jpg'
    Image = cv2.imread(ImgPath)
    print(predict(Model,Image,Device))

---

以上就是全部代码和实现过程，图就不放了，懒得搞了