Question

我们可以在Tensorflow对象检测API类中对检测到的对象进行计数吗？假设您有car和bicyle这两个类，我想将每个对象类明智地算出；，Carcount：5自行车数：3。

Answer 1

实现此目标的一种方法是执行以下操作

boxes = detection_graph.get_tensor_by_name('detection_boxes:0')
scores = detection_graph.get_tensor_by_name('detection_scores:0')
classes = detection_graph.get_tensor_by_name('detection_classes:0')
num_detections = detection_graph.get_tensor_by_name('num_detections:0')
(boxes, scores, classes, num_detections) = sess.run([boxes, scores, classes, num_detections],feed_dict={image_tensor: image_np_expanded})

final_score = np.squeeze(scores)    
car_count = 0

for i in range(100):
    if final_score[i] > 0.5:
        detected_class = int(classes[0][i])
        if detected_class == 1:
            car_count += 1

在这里，根据您在label_map.pbtxt文件中指定的值映射了类名和编号。我希望这能够帮到你。如果您遇到任何问题，请告诉我

类明智的Tensorflow对象检测计数

1 个答案: