使用TensorFlow对象检测提取输出分数、类、id和框

据此,。 我的是； 让我们假设有一张图片包含3只猫、2只狗和1只鸟。 在检测到整个对象后，我们如何获得分离的6个对象的xmin-ymin-xmax-ymax值。

在这些行之后

    (boxes, scores, classes, num_detections) = sess.run(
          [boxes, scores, classes, num_detections],
          feed_dict={image_tensor: image_np_expanded})

您可以检索需要查找的信息

    boxes, scores, classes, num_detections

在这几行之后

    (boxes, scores, classes, num_detections) = sess.run(
          [boxes, scores, classes, num_detections],
          feed_dict={image_tensor: image_np_expanded})

您可以检索需要查找的信息

    boxes, scores, classes, num_detections

---

在Python中，它看起来像

    # this loop Counting the Objects found from highest to lowest %, Default is 100Results. Only > x% get counted
        scores = output_dict['detection_scores'] as example
        boxes = output_dict['detection_boxes'] as example
        classes = output_dict['detection_classes'] as example
        count=0  
        xmin=[]
        xmax=[]
        ymin=[]
        ymax=[] 
        classlist=[]         
        for s in range (100):
            if scores is None or scores [s] > 0.5:
                count = count + 1
        for i in range (count):
            position = np.squeeze(boxes[0][i])
            (xmin, xmax, ymin, ymax) = (position[1]*im_width, position[3]*im_width, position[0]*im_height, position[2]*im_height)
        xmin.append(xmin)
        xmax.append(xmax)
        ymin.append(ymin)
        ymax.append(ymax)
        classlist.append(classes[i])

列表按得分从高到低排序。

---

很抱歉，我是新手。

在Python中，它看起来像

    # this loop Counting the Objects found from highest to lowest %, Default is 100Results. Only > x% get counted
        scores = output_dict['detection_scores'] as example
        boxes = output_dict['detection_boxes'] as example
        classes = output_dict['detection_classes'] as example
        count=0  
        xmin=[]
        xmax=[]
        ymin=[]
        ymax=[] 
        classlist=[]         
        for s in range (100):
            if scores is None or scores [s] > 0.5:
                count = count + 1
        for i in range (count):
            position = np.squeeze(boxes[0][i])
            (xmin, xmax, ymin, ymax) = (position[1]*im_width, position[3]*im_width, position[0]*im_height, position[2]*im_height)
        xmin.append(xmin)
        xmax.append(xmax)
        ymin.append(ymin)
        ymax.append(ymax)
        classlist.append(classes[i])

列表按得分从高到低排序。 对不起，我是新手