opencv - 使用 DummyData 层 : "Can' t create layer "DummyData1" of type "DummyData1" in function 'getLayerInstance' "加载 Caffe 模型时出错

Question

我将 Tensorflow 模型转换为 Caffe 模型( tf_resnet.prototxt )，现在我正在尝试运行对象检测计数算法( code )。但是，我收到以下错误:

[INFO] loading model...
[INFO] opening video file...
Traceback (most recent call last):
  File "people_counter.py", line 132, in <module>
    detections = net.forward()
cv2.error: OpenCV(4.2.0) /io/opencv/modules/dnn/src/dnn.cpp:562: error: (-2:Unspecified error) Can't create layer "DummyData1" of type "DummyData" in function 'getLayerInstance'

This Stackoverflow 上的类似问题建议更改 type: "DummyData1"至 type: "Input"在以下代码中:

layer {
  name: "DummyData1"
  type: "DummyData1"
  top: "DummyData1"
  dummy_data_param {
    shape {
      dim: 1
      dim: 64
      dim: 150
      dim: 150
    }
  }
}

但是，在执行此操作时，我收到另一个错误:

Traceback (most recent call last):
  File "people_counter.py", line 132, in <module>
    detections = net.forward()
cv2.error: OpenCV(4.2.0) /io/opencv/modules/dnn/src/dnn.cpp:2709: error: (-215:Assertion failed) inp.total() in function 'allocateLayers''

有关如何解决此问题的任何建议？

Answer 1

你可以试试这个方法；改变这个:

layer {
  name: "DummyData1"
  type: "DummyData1"
  top: "DummyData1"
  dummy_data_param {
    shape {
      dim: 1
      dim: 64
      dim: 150
      dim: 150
    }
  }
}

有了这个:

layer {
  name: "DummyData1"
  type: "Input"
  top: "DummyData1"
  input_param {
    shape {
      dim: 1
      dim: 64
      dim: 150
      dim: 150
    }
  }
}

然后在脚本中，而不是使用这些行:

blob = cv2.dnn.blobFromImage(frame, 0.007843, (W, H), 127.5)
net.setInput(blob)
detections = net.forward()

试试这些:

blob = cv2.dnn.blobFromImage(frame, 0.007843, (W, H), 127.5)
net.setInput(blob, "Placeholder")
net.setInput(np.zeros((1, 64, 150, 150)), "DummyData1")
detections = net.forward()

我不知道结果是否会好，但这至少应该解决异常。
为了获得良好的结果，您必须找出该层中的数据类型，因为在原始模型中，“DummyData1”可能是某种生成器或常量矩阵。