Toybrick

标题: deprecated caffe input usage, please change it to input layer [打印本页]

作者: chuyee 时间: 2019-2-13 08:57
标题: deprecated caffe input usage, please change it to input layer
本帖最后由 chuyee 于 2019-2-27 01:49 编辑

$ python test.py
/usr/lib64/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.
  from ._conv import register_converters as _register_converters
--> config model
done
--> Loading model
Catch exception when loading caffe model: ./pose_deploy_linevec.prototxt!
Traceback (most recent call last):
  File "/usr/local/lib64/python3.6/site-packages/rknn/api/rknn.py", line 75, in load_caffe
ret = self.rknn_base.load_caffe(model, proto, blobs)
  File "rknn/api/redirect_stdout.py", line 67, in rknn.api.redirect_stdout.redirect_stdout.func_wrapper
  File "rknn/api/redirect_stdout.py", line 68, in rknn.api.redirect_stdout.redirect_stdout.func_wrapper
  File "rknn/api/rknn_base.py", line 240, in rknn.api.rknn_base.RKNNBase.load_caffe
  File "rknn/base/rknnlib/converter/caffeloader.py", line 963, in rknn.base.rknnlib.converter.caffeloader.CaffeLoader.load
  File "rknn/base/rknnlib/converter/caffeloader.py", line 745, in rknn.base.rknnlib.converter.caffeloader.CaffeLoader.parse_net_param
  File "rknn/base/rknnlib/rknnlog.py", line 105, in rknn.base.rknnlib.rknnlog.rknnLog.e
ValueError: Deprecated caffe input usage, please change it to input layer.

Which part of the input is deprecated? How to change the input layer?

作者: yhc 时间: 2019-2-13 15:39
prototxt格式是旧的，可以按照下面格式修改

name: "MOBILENET_V2"

#  transform_param {

#    scale: 0.017

#    mirror: false

#    crop_size: 224

#    mean_value: [103.94,116.78,123.68]

#  }

layer {

  name: "data"

  type: "Input"

  top: "data"

  input_param {

    shape {

      dim: 1

      dim: 3

      dim: 224

      dim: 224

    }

  }

}
复制代码

作者: chuyee 时间: 2019-2-15 09:34

yhc 发表于 2019-2-13 15:39
prototxt格式是旧的，可以按照下面格式修改

Thanks! I modified the input layer according to your example. The "input is deprecated" error goes away now. However I encountered a new problem as below. Could you please help to check what might be wrong with it?

--> Loading model
Catch exception when loading caffe model: ./pose_modified.prototxt!
Traceback (most recent call last):
  File "/usr/local/lib64/python3.6/site-packages/rknn/api/rknn.py", line 75, in load_caffe
ret = self.rknn_base.load_caffe(model, proto, blobs)
  File "rknn/api/redirect_stdout.py", line 67, in rknn.api.redirect_stdout.redirect_stdout.func_wrapper
  File "rknn/api/redirect_stdout.py", line 68, in rknn.api.redirect_stdout.redirect_stdout.func_wrapper
  File "rknn/api/rknn_base.py", line 243, in rknn.api.rknn_base.RKNNBase.load_caffe
  File "rknn/base/rknnlib/rknnnetbuilder.py", line 114, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build
  File "rknn/base/rknnlib/rknnnetbuilder.py", line 132, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
  File "rknn/base/rknnlib/rknnnetbuilder.py", line 132, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
  File "rknn/base/rknnlib/rknnnetbuilder.py", line 132, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
  [Previous line repeated 104 more times]
  File "rknn/base/rknnlib/rknnnetbuilder.py", line 138, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
  File "rknn/base/rknnlib/layer/rknnlayer.py", line 247, in rknn.base.rknnlib.layer.rknnlayer.rknnLayer.compute_shape
  File "rknn/base/rknnlib/layer/convolution.py", line 65, in rknn.base.rknnlib.layer.convolution.Convolution.compute_out_shape
AttributeError: 'NoneType' object has no attribute 'format'

作者: chuyee 时间: 2019-2-15 09:38
The model prototxt is https://github.com/CMU-Perceptua ... oy_linevec.prototxt

I replaced the first 5 lines with below according to your suggestion.

name: "pose"
layer {
  name: "data"
  type: "Input"
  top: "data"
  input_param {
shape {
   dim: 1
   dim: 3
   dim: 224
   dim: 224
}
  }
}

作者: 慢慢的大米饭 时间: 2019-2-15 15:57

chuyee 发表于 2019-2-15 09:38
The model prototxt is https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/models/po ...

请问解决了吗

作者: chuyee 时间: 2019-2-15 16:42

慢慢的大米饭发表于 2019-2-15 15:57
请问解决了吗

No, please help!

作者: chuyee 时间: 2019-2-15 16:47
This is the verbose output from 0.9.8 toolkit.

--> config model
done
--> Loading model
I Set caffe proto to caffe
I Load caffe model ./pose_modified.prototxt
I Parsing net parameters ...
D import clients finished
I Parsing layer parameters ...
D Convert layer data
D Convert layer conv1_1
D Convert layer relu1_1
D Convert layer conv1_2
D Convert layer relu1_2
D Convert layer pool1_stage1
D Convert layer conv2_1
D Convert layer relu2_1
D Convert layer conv2_2
D Convert layer relu2_2
D Convert layer pool2_stage1
D Convert layer conv3_1
D Convert layer relu3_1
D Convert layer conv3_2
D Convert layer relu3_2
D Convert layer conv3_3
D Convert layer relu3_3
D Convert layer conv3_4
D Convert layer relu3_4
D Convert layer pool3_stage1
D Convert layer conv4_1
D Convert layer relu4_1
D Convert layer conv4_2
D Convert layer relu4_2
D Convert layer conv4_3_CPM
D Convert layer relu4_3_CPM
D Convert layer conv4_4_CPM
D Convert layer relu4_4_CPM
D Convert layer conv5_1_CPM_L1
D Convert layer relu5_1_CPM_L1
D Convert layer conv5_1_CPM_L2
D Convert layer relu5_1_CPM_L2
D Convert layer conv5_2_CPM_L1
D Convert layer relu5_2_CPM_L1
D Convert layer conv5_2_CPM_L2
D Convert layer relu5_2_CPM_L2
D Convert layer conv5_3_CPM_L1
D Convert layer relu5_3_CPM_L1
D Convert layer conv5_3_CPM_L2
D Convert layer relu5_3_CPM_L2
D Convert layer conv5_4_CPM_L1
D Convert layer relu5_4_CPM_L1
D Convert layer conv5_4_CPM_L2
D Convert layer relu5_4_CPM_L2
D Convert layer conv5_5_CPM_L1
D Convert layer conv5_5_CPM_L2
D Convert layer concat_stage2
D Convert layer Mconv1_stage2_L1
D Convert layer Mrelu1_stage2_L1
D Convert layer Mconv1_stage2_L2
D Convert layer Mrelu1_stage2_L2
D Convert layer Mconv2_stage2_L1
D Convert layer Mrelu2_stage2_L1
D Convert layer Mconv2_stage2_L2
D Convert layer Mrelu2_stage2_L2
D Convert layer Mconv3_stage2_L1
D Convert layer Mrelu3_stage2_L1
D Convert layer Mconv3_stage2_L2
D Convert layer Mrelu3_stage2_L2
D Convert layer Mconv4_stage2_L1
D Convert layer Mrelu4_stage2_L1
D Convert layer Mconv4_stage2_L2
D Convert layer Mrelu4_stage2_L2
D Convert layer Mconv5_stage2_L1
D Convert layer Mrelu5_stage2_L1
D Convert layer Mconv5_stage2_L2
D Convert layer Mrelu5_stage2_L2
D Convert layer Mconv6_stage2_L1
D Convert layer Mrelu6_stage2_L1
D Convert layer Mconv6_stage2_L2
D Convert layer Mrelu6_stage2_L2
D Convert layer Mconv7_stage2_L1
D Convert layer Mconv7_stage2_L2
D Convert layer concat_stage3
D Convert layer Mconv1_stage3_L1
D Convert layer Mrelu1_stage3_L1
D Convert layer Mconv1_stage3_L2
D Convert layer Mrelu1_stage3_L2
D Convert layer Mconv2_stage3_L1
D Convert layer Mrelu2_stage3_L1
D Convert layer Mconv2_stage3_L2
D Convert layer Mrelu2_stage3_L2
D Convert layer Mconv3_stage3_L1
D Convert layer Mrelu3_stage3_L1
D Convert layer Mconv3_stage3_L2
D Convert layer Mrelu3_stage3_L2
D Convert layer Mconv4_stage3_L1
D Convert layer Mrelu4_stage3_L1
D Convert layer Mconv4_stage3_L2
D Convert layer Mrelu4_stage3_L2
D Convert layer Mconv5_stage3_L1
D Convert layer Mrelu5_stage3_L1
D Convert layer Mconv5_stage3_L2
D Convert layer Mrelu5_stage3_L2
D Convert layer Mconv6_stage3_L1
D Convert layer Mrelu6_stage3_L1
D Convert layer Mconv6_stage3_L2
D Convert layer Mrelu6_stage3_L2
D Convert layer Mconv7_stage3_L1
D Convert layer Mconv7_stage3_L2
D Convert layer concat_stage4
D Convert layer Mconv1_stage4_L1
D Convert layer Mrelu1_stage4_L1
D Convert layer Mconv1_stage4_L2
D Convert layer Mrelu1_stage4_L2
D Convert layer Mconv2_stage4_L1
D Convert layer Mrelu2_stage4_L1
D Convert layer Mconv2_stage4_L2
D Convert layer Mrelu2_stage4_L2
D Convert layer Mconv3_stage4_L1
D Convert layer Mrelu3_stage4_L1
D Convert layer Mconv3_stage4_L2
D Convert layer Mrelu3_stage4_L2
D Convert layer Mconv4_stage4_L1
D Convert layer Mrelu4_stage4_L1
D Convert layer Mconv4_stage4_L2
D Convert layer Mrelu4_stage4_L2
D Convert layer Mconv5_stage4_L1
D Convert layer Mrelu5_stage4_L1
D Convert layer Mconv5_stage4_L2
D Convert layer Mrelu5_stage4_L2
D Convert layer Mconv6_stage4_L1
D Convert layer Mrelu6_stage4_L1
D Convert layer Mconv6_stage4_L2
D Convert layer Mrelu6_stage4_L2
D Convert layer Mconv7_stage4_L1
D Convert layer Mconv7_stage4_L2
D Convert layer concat_stage5
D Convert layer Mconv1_stage5_L1
D Convert layer Mrelu1_stage5_L1
D Convert layer Mconv1_stage5_L2
D Convert layer Mrelu1_stage5_L2
D Convert layer Mconv2_stage5_L1
D Convert layer Mrelu2_stage5_L1
D Convert layer Mconv2_stage5_L2
D Convert layer Mrelu2_stage5_L2
D Convert layer Mconv3_stage5_L1
D Convert layer Mrelu3_stage5_L1
D Convert layer Mconv3_stage5_L2
D Convert layer Mrelu3_stage5_L2
D Convert layer Mconv4_stage5_L1
D Convert layer Mrelu4_stage5_L1
D Convert layer Mconv4_stage5_L2
D Convert layer Mrelu4_stage5_L2
D Convert layer Mconv5_stage5_L1
D Convert layer Mrelu5_stage5_L1
D Convert layer Mconv5_stage5_L2
D Convert layer Mrelu5_stage5_L2
D Convert layer Mconv6_stage5_L1
D Convert layer Mrelu6_stage5_L1
D Convert layer Mconv6_stage5_L2
D Convert layer Mrelu6_stage5_L2
D Convert layer Mconv7_stage5_L1
D Convert layer Mconv7_stage5_L2
D Convert layer concat_stage6
D Convert layer Mconv1_stage6_L1
D Convert layer Mrelu1_stage6_L1
D Convert layer Mconv1_stage6_L2
D Convert layer Mrelu1_stage6_L2
D Convert layer Mconv2_stage6_L1
D Convert layer Mrelu2_stage6_L1
D Convert layer Mconv2_stage6_L2
D Convert layer Mrelu2_stage6_L2
D Convert layer Mconv3_stage6_L1
D Convert layer Mrelu3_stage6_L1
D Convert layer Mconv3_stage6_L2
D Convert layer Mrelu3_stage6_L2
D Convert layer Mconv4_stage6_L1
D Convert layer Mrelu4_stage6_L1
D Convert layer Mconv4_stage6_L2
D Convert layer Mrelu4_stage6_L2
D Convert layer Mconv5_stage6_L1
D Convert layer Mrelu5_stage6_L1
D Convert layer Mconv5_stage6_L2
D Convert layer Mrelu5_stage6_L2
D Convert layer Mconv6_stage6_L1
D Convert layer Mrelu6_stage6_L1
D Convert layer Mconv6_stage6_L2
D Convert layer Mrelu6_stage6_L2
D Convert layer Mconv7_stage6_L1
D Convert layer Mconv7_stage6_L2
D Convert layer concat_stage7
I Parsing connections ...
D Connect: data_0,0 to output_182,0,
D Connect: conv1_1_1,0 to relu1_1_2,0
D Connect: relu1_1_2,0 to conv1_2_3,0
D Connect: conv1_2_3,0 to relu1_2_4,0
D Connect: relu1_2_4,0 to pool1_stage1_5,0
D Connect: pool1_stage1_5,0 to conv2_1_6,0
D Connect: conv2_1_6,0 to relu2_1_7,0
D Connect: relu2_1_7,0 to conv2_2_8,0
D Connect: conv2_2_8,0 to relu2_2_9,0
D Connect: relu2_2_9,0 to pool2_stage1_10,0
D Connect: pool2_stage1_10,0 to conv3_1_11,0
D Connect: conv3_1_11,0 to relu3_1_12,0
D Connect: relu3_1_12,0 to conv3_2_13,0
D Connect: conv3_2_13,0 to relu3_2_14,0
D Connect: relu3_2_14,0 to conv3_3_15,0
D Connect: conv3_3_15,0 to relu3_3_16,0
D Connect: relu3_3_16,0 to conv3_4_17,0
D Connect: conv3_4_17,0 to relu3_4_18,0
D Connect: relu3_4_18,0 to pool3_stage1_19,0
D Connect: pool3_stage1_19,0 to conv4_1_20,0
D Connect: conv4_1_20,0 to relu4_1_21,0
D Connect: relu4_1_21,0 to conv4_2_22,0
D Connect: conv4_2_22,0 to relu4_2_23,0
D Connect: relu4_2_23,0 to conv4_3_CPM_24,0
D Connect: conv4_3_CPM_24,0 to relu4_3_CPM_25,0
D Connect: relu4_3_CPM_25,0 to conv4_4_CPM_26,0
D Connect: conv4_4_CPM_26,0 to relu4_4_CPM_27,0
D Connect: relu4_4_CPM_27,0 to conv5_1_CPM_L1_28,0
D Connect: relu4_4_CPM_27,0 to conv5_1_CPM_L2_30,0
D Connect: relu4_4_CPM_27,0 to concat_stage2_46,2
D Connect: relu4_4_CPM_27,0 to concat_stage3_73,2
D Connect: relu4_4_CPM_27,0 to concat_stage4_100,2
D Connect: relu4_4_CPM_27,0 to concat_stage5_127,2
D Connect: relu4_4_CPM_27,0 to concat_stage6_154,2
D Connect: conv5_1_CPM_L1_28,0 to relu5_1_CPM_L1_29,0
D Connect: relu5_1_CPM_L1_29,0 to conv5_2_CPM_L1_32,0
D Connect: conv5_1_CPM_L2_30,0 to relu5_1_CPM_L2_31,0
D Connect: relu5_1_CPM_L2_31,0 to conv5_2_CPM_L2_34,0
D Connect: conv5_2_CPM_L1_32,0 to relu5_2_CPM_L1_33,0
D Connect: relu5_2_CPM_L1_33,0 to conv5_3_CPM_L1_36,0
D Connect: conv5_2_CPM_L2_34,0 to relu5_2_CPM_L2_35,0
D Connect: relu5_2_CPM_L2_35,0 to conv5_3_CPM_L2_38,0
D Connect: conv5_3_CPM_L1_36,0 to relu5_3_CPM_L1_37,0
D Connect: relu5_3_CPM_L1_37,0 to conv5_4_CPM_L1_40,0
D Connect: conv5_3_CPM_L2_38,0 to relu5_3_CPM_L2_39,0
D Connect: relu5_3_CPM_L2_39,0 to conv5_4_CPM_L2_42,0
D Connect: conv5_4_CPM_L1_40,0 to relu5_4_CPM_L1_41,0
D Connect: relu5_4_CPM_L1_41,0 to conv5_5_CPM_L1_44,0
D Connect: conv5_4_CPM_L2_42,0 to relu5_4_CPM_L2_43,0
D Connect: relu5_4_CPM_L2_43,0 to conv5_5_CPM_L2_45,0
D Connect: conv5_5_CPM_L1_44,0 to concat_stage2_46,0
D Connect: conv5_5_CPM_L2_45,0 to concat_stage2_46,1
D Connect: concat_stage2_46,0 to Mconv1_stage2_L1_47,0
D Connect: concat_stage2_46,0 to Mconv1_stage2_L2_49,0
D Connect: Mconv1_stage2_L1_47,0 to Mrelu1_stage2_L1_48,0
D Connect: Mrelu1_stage2_L1_48,0 to Mconv2_stage2_L1_51,0
D Connect: Mconv1_stage2_L2_49,0 to Mrelu1_stage2_L2_50,0
D Connect: Mrelu1_stage2_L2_50,0 to Mconv2_stage2_L2_53,0
D Connect: Mconv2_stage2_L1_51,0 to Mrelu2_stage2_L1_52,0
D Connect: Mrelu2_stage2_L1_52,0 to Mconv3_stage2_L1_55,0
D Connect: Mconv2_stage2_L2_53,0 to Mrelu2_stage2_L2_54,0
D Connect: Mrelu2_stage2_L2_54,0 to Mconv3_stage2_L2_57,0
D Connect: Mconv3_stage2_L1_55,0 to Mrelu3_stage2_L1_56,0
D Connect: Mrelu3_stage2_L1_56,0 to Mconv4_stage2_L1_59,0
D Connect: Mconv3_stage2_L2_57,0 to Mrelu3_stage2_L2_58,0
D Connect: Mrelu3_stage2_L2_58,0 to Mconv4_stage2_L2_61,0
D Connect: Mconv4_stage2_L1_59,0 to Mrelu4_stage2_L1_60,0
D Connect: Mrelu4_stage2_L1_60,0 to Mconv5_stage2_L1_63,0
D Connect: Mconv4_stage2_L2_61,0 to Mrelu4_stage2_L2_62,0
D Connect: Mrelu4_stage2_L2_62,0 to Mconv5_stage2_L2_65,0
D Connect: Mconv5_stage2_L1_63,0 to Mrelu5_stage2_L1_64,0
D Connect: Mrelu5_stage2_L1_64,0 to Mconv6_stage2_L1_67,0
D Connect: Mconv5_stage2_L2_65,0 to Mrelu5_stage2_L2_66,0
D Connect: Mrelu5_stage2_L2_66,0 to Mconv6_stage2_L2_69,0
D Connect: Mconv6_stage2_L1_67,0 to Mrelu6_stage2_L1_68,0
D Connect: Mrelu6_stage2_L1_68,0 to Mconv7_stage2_L1_71,0
D Connect: Mconv6_stage2_L2_69,0 to Mrelu6_stage2_L2_70,0
D Connect: Mrelu6_stage2_L2_70,0 to Mconv7_stage2_L2_72,0
D Connect: Mconv7_stage2_L1_71,0 to concat_stage3_73,0
D Connect: Mconv7_stage2_L2_72,0 to concat_stage3_73,1
D Connect: concat_stage3_73,0 to Mconv1_stage3_L1_74,0
D Connect: concat_stage3_73,0 to Mconv1_stage3_L2_76,0
D Connect: Mconv1_stage3_L1_74,0 to Mrelu1_stage3_L1_75,0
D Connect: Mrelu1_stage3_L1_75,0 to Mconv2_stage3_L1_78,0
D Connect: Mconv1_stage3_L2_76,0 to Mrelu1_stage3_L2_77,0
D Connect: Mrelu1_stage3_L2_77,0 to Mconv2_stage3_L2_80,0
D Connect: Mconv2_stage3_L1_78,0 to Mrelu2_stage3_L1_79,0
D Connect: Mrelu2_stage3_L1_79,0 to Mconv3_stage3_L1_82,0
D Connect: Mconv2_stage3_L2_80,0 to Mrelu2_stage3_L2_81,0
D Connect: Mrelu2_stage3_L2_81,0 to Mconv3_stage3_L2_84,0
D Connect: Mconv3_stage3_L1_82,0 to Mrelu3_stage3_L1_83,0
D Connect: Mrelu3_stage3_L1_83,0 to Mconv4_stage3_L1_86,0
D Connect: Mconv3_stage3_L2_84,0 to Mrelu3_stage3_L2_85,0
D Connect: Mrelu3_stage3_L2_85,0 to Mconv4_stage3_L2_88,0
D Connect: Mconv4_stage3_L1_86,0 to Mrelu4_stage3_L1_87,0
D Connect: Mrelu4_stage3_L1_87,0 to Mconv5_stage3_L1_90,0
D Connect: Mconv4_stage3_L2_88,0 to Mrelu4_stage3_L2_89,0
D Connect: Mrelu4_stage3_L2_89,0 to Mconv5_stage3_L2_92,0
D Connect: Mconv5_stage3_L1_90,0 to Mrelu5_stage3_L1_91,0
D Connect: Mrelu5_stage3_L1_91,0 to Mconv6_stage3_L1_94,0
D Connect: Mconv5_stage3_L2_92,0 to Mrelu5_stage3_L2_93,0
D Connect: Mrelu5_stage3_L2_93,0 to Mconv6_stage3_L2_96,0
D Connect: Mconv6_stage3_L1_94,0 to Mrelu6_stage3_L1_95,0
D Connect: Mrelu6_stage3_L1_95,0 to Mconv7_stage3_L1_98,0
D Connect: Mconv6_stage3_L2_96,0 to Mrelu6_stage3_L2_97,0
D Connect: Mrelu6_stage3_L2_97,0 to Mconv7_stage3_L2_99,0
D Connect: Mconv7_stage3_L1_98,0 to concat_stage4_100,0
D Connect: Mconv7_stage3_L2_99,0 to concat_stage4_100,1
D Connect: concat_stage4_100,0 to Mconv1_stage4_L1_101,0
D Connect: concat_stage4_100,0 to Mconv1_stage4_L2_103,0
D Connect: Mconv1_stage4_L1_101,0 to Mrelu1_stage4_L1_102,0
D Connect: Mrelu1_stage4_L1_102,0 to Mconv2_stage4_L1_105,0
D Connect: Mconv1_stage4_L2_103,0 to Mrelu1_stage4_L2_104,0
D Connect: Mrelu1_stage4_L2_104,0 to Mconv2_stage4_L2_107,0
D Connect: Mconv2_stage4_L1_105,0 to Mrelu2_stage4_L1_106,0
D Connect: Mrelu2_stage4_L1_106,0 to Mconv3_stage4_L1_109,0
D Connect: Mconv2_stage4_L2_107,0 to Mrelu2_stage4_L2_108,0
D Connect: Mrelu2_stage4_L2_108,0 to Mconv3_stage4_L2_111,0
D Connect: Mconv3_stage4_L1_109,0 to Mrelu3_stage4_L1_110,0
D Connect: Mrelu3_stage4_L1_110,0 to Mconv4_stage4_L1_113,0
D Connect: Mconv3_stage4_L2_111,0 to Mrelu3_stage4_L2_112,0
D Connect: Mrelu3_stage4_L2_112,0 to Mconv4_stage4_L2_115,0
D Connect: Mconv4_stage4_L1_113,0 to Mrelu4_stage4_L1_114,0
D Connect: Mrelu4_stage4_L1_114,0 to Mconv5_stage4_L1_117,0
D Connect: Mconv4_stage4_L2_115,0 to Mrelu4_stage4_L2_116,0
D Connect: Mrelu4_stage4_L2_116,0 to Mconv5_stage4_L2_119,0
D Connect: Mconv5_stage4_L1_117,0 to Mrelu5_stage4_L1_118,0
D Connect: Mrelu5_stage4_L1_118,0 to Mconv6_stage4_L1_121,0
D Connect: Mconv5_stage4_L2_119,0 to Mrelu5_stage4_L2_120,0
D Connect: Mrelu5_stage4_L2_120,0 to Mconv6_stage4_L2_123,0
D Connect: Mconv6_stage4_L1_121,0 to Mrelu6_stage4_L1_122,0
D Connect: Mrelu6_stage4_L1_122,0 to Mconv7_stage4_L1_125,0
D Connect: Mconv6_stage4_L2_123,0 to Mrelu6_stage4_L2_124,0
D Connect: Mrelu6_stage4_L2_124,0 to Mconv7_stage4_L2_126,0
D Connect: Mconv7_stage4_L1_125,0 to concat_stage5_127,0
D Connect: Mconv7_stage4_L2_126,0 to concat_stage5_127,1
D Connect: concat_stage5_127,0 to Mconv1_stage5_L1_128,0
D Connect: concat_stage5_127,0 to Mconv1_stage5_L2_130,0
D Connect: Mconv1_stage5_L1_128,0 to Mrelu1_stage5_L1_129,0
D Connect: Mrelu1_stage5_L1_129,0 to Mconv2_stage5_L1_132,0
D Connect: Mconv1_stage5_L2_130,0 to Mrelu1_stage5_L2_131,0
D Connect: Mrelu1_stage5_L2_131,0 to Mconv2_stage5_L2_134,0
D Connect: Mconv2_stage5_L1_132,0 to Mrelu2_stage5_L1_133,0
D Connect: Mrelu2_stage5_L1_133,0 to Mconv3_stage5_L1_136,0
D Connect: Mconv2_stage5_L2_134,0 to Mrelu2_stage5_L2_135,0
D Connect: Mrelu2_stage5_L2_135,0 to Mconv3_stage5_L2_138,0
D Connect: Mconv3_stage5_L1_136,0 to Mrelu3_stage5_L1_137,0
D Connect: Mrelu3_stage5_L1_137,0 to Mconv4_stage5_L1_140,0
D Connect: Mconv3_stage5_L2_138,0 to Mrelu3_stage5_L2_139,0
D Connect: Mrelu3_stage5_L2_139,0 to Mconv4_stage5_L2_142,0
D Connect: Mconv4_stage5_L1_140,0 to Mrelu4_stage5_L1_141,0
D Connect: Mrelu4_stage5_L1_141,0 to Mconv5_stage5_L1_144,0
D Connect: Mconv4_stage5_L2_142,0 to Mrelu4_stage5_L2_143,0
D Connect: Mrelu4_stage5_L2_143,0 to Mconv5_stage5_L2_146,0
D Connect: Mconv5_stage5_L1_144,0 to Mrelu5_stage5_L1_145,0
D Connect: Mrelu5_stage5_L1_145,0 to Mconv6_stage5_L1_148,0
D Connect: Mconv5_stage5_L2_146,0 to Mrelu5_stage5_L2_147,0
D Connect: Mrelu5_stage5_L2_147,0 to Mconv6_stage5_L2_150,0
D Connect: Mconv6_stage5_L1_148,0 to Mrelu6_stage5_L1_149,0
D Connect: Mrelu6_stage5_L1_149,0 to Mconv7_stage5_L1_152,0
D Connect: Mconv6_stage5_L2_150,0 to Mrelu6_stage5_L2_151,0
D Connect: Mrelu6_stage5_L2_151,0 to Mconv7_stage5_L2_153,0
D Connect: Mconv7_stage5_L1_152,0 to concat_stage6_154,0
D Connect: Mconv7_stage5_L2_153,0 to concat_stage6_154,1
D Connect: concat_stage6_154,0 to Mconv1_stage6_L1_155,0
D Connect: concat_stage6_154,0 to Mconv1_stage6_L2_157,0
D Connect: Mconv1_stage6_L1_155,0 to Mrelu1_stage6_L1_156,0
D Connect: Mrelu1_stage6_L1_156,0 to Mconv2_stage6_L1_159,0
D Connect: Mconv1_stage6_L2_157,0 to Mrelu1_stage6_L2_158,0
D Connect: Mrelu1_stage6_L2_158,0 to Mconv2_stage6_L2_161,0
D Connect: Mconv2_stage6_L1_159,0 to Mrelu2_stage6_L1_160,0
D Connect: Mrelu2_stage6_L1_160,0 to Mconv3_stage6_L1_163,0
D Connect: Mconv2_stage6_L2_161,0 to Mrelu2_stage6_L2_162,0
D Connect: Mrelu2_stage6_L2_162,0 to Mconv3_stage6_L2_165,0
D Connect: Mconv3_stage6_L1_163,0 to Mrelu3_stage6_L1_164,0
D Connect: Mrelu3_stage6_L1_164,0 to Mconv4_stage6_L1_167,0
D Connect: Mconv3_stage6_L2_165,0 to Mrelu3_stage6_L2_166,0
D Connect: Mrelu3_stage6_L2_166,0 to Mconv4_stage6_L2_169,0
D Connect: Mconv4_stage6_L1_167,0 to Mrelu4_stage6_L1_168,0
D Connect: Mrelu4_stage6_L1_168,0 to Mconv5_stage6_L1_171,0
D Connect: Mconv4_stage6_L2_169,0 to Mrelu4_stage6_L2_170,0
D Connect: Mrelu4_stage6_L2_170,0 to Mconv5_stage6_L2_173,0
D Connect: Mconv5_stage6_L1_171,0 to Mrelu5_stage6_L1_172,0
D Connect: Mrelu5_stage6_L1_172,0 to Mconv6_stage6_L1_175,0
D Connect: Mconv5_stage6_L2_173,0 to Mrelu5_stage6_L2_174,0
D Connect: Mrelu5_stage6_L2_174,0 to Mconv6_stage6_L2_177,0
D Connect: Mconv6_stage6_L1_175,0 to Mrelu6_stage6_L1_176,0
D Connect: Mrelu6_stage6_L1_176,0 to Mconv7_stage6_L1_179,0
D Connect: Mconv6_stage6_L2_177,0 to Mrelu6_stage6_L2_178,0
D Connect: Mrelu6_stage6_L2_178,0 to Mconv7_stage6_L2_180,0
D Connect: Mconv7_stage6_L1_179,0 to concat_stage7_181,1
D Connect: Mconv7_stage6_L2_180,0 to concat_stage7_181,0
D Connect: concat_stage7_181,0 to output_183,0,
I Load net complete.
D Process data_0 ...
D RKNN output shape(input): (0 224 224 3)
D Process output_182 ...
D RKNN output shape(output): (0 224 224 3)
D Process conv1_1_1 ...
E Catch exception when loading caffe model: ./pose_modified.prototxt!
T Traceback (most recent call last):
T File "rknn/api/rknn_base.py", line 260, in rknn.api.rknn_base.RKNNBase.load_caffe
T File "rknn/base/rknnlib/rknnnetbuilder.py", line 116, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build
T File "rknn/base/rknnlib/rknnnetbuilder.py", line 134, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
T File "rknn/base/rknnlib/rknnnetbuilder.py", line 134, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
T File "rknn/base/rknnlib/rknnnetbuilder.py", line 134, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
T [Previous line repeated 104 more times]
T File "rknn/base/rknnlib/rknnnetbuilder.py", line 140, in rknn.base.rknnlib.rknnnetbuilder.rknnNetBuilder.build_layer
T File "rknn/base/rknnlib/layer/rknnlayer.py", line 247, in rknn.base.rknnlib.layer.rknnlayer.rknnLayer.compute_shape
T File "rknn/base/rknnlib/layer/convolution.py", line 65, in rknn.base.rknnlib.layer.convolution.Convolution.compute_out_shape
T AttributeError: 'NoneType' object has no attribute 'format'
Load mobilenet_v2 failed! Ret = -1

作者: 慢慢的大米饭 时间: 2019-2-16 10:22

chuyee 发表于 2019-2-15 16:47
This is the verbose output from 0.9.8 toolkit.

--> config model

我报的错跟你不一样我是有一个layer 未知你可以看下我的帖子官方的人一点信都没失望啊

作者: yhc 时间: 2019-2-18 09:51
这个看layer都有支持的，caffemodel文件请问是哪里找到的呢

作者: chuyee 时间: 2019-2-19 15:35

yhc 发表于 2019-2-18 09:51
这个看layer都有支持的，caffemodel文件请问是哪里找到的呢

http://posefs1.perception.cs.cmu ... r_440000.caffemodel

This is the standard OpenPose coco caffe pose model.

作者: yhc 时间: 2019-2-19 17:53
protetxt请改成这样

layer {

  name: "data"

  type: "Input"

  top: "image"

  input_param {

    shape {

      dim: 1

      dim: 3

      dim: 160

      dim: 160

    }

  }

}
复制代码

作者: chuyee 时间: 2019-2-21 06:45

yhc 发表于 2019-2-19 17:53
protetxt请改成这样

可以工作了，谢谢！

作者: kitedream 时间: 2019-3-13 20:13
本帖最后由 kitedream 于 2019-3-14 17:04 编辑

抱歉，我漏掉rknn的一个初始化部分。

作者: kitedream 时间: 2019-3-13 22:13
本帖最后由 kitedream 于 2019-3-14 17:03 编辑

抱歉，我初始化部分没有做好

作者: LSC 时间: 2019-12-18 15:46

chuyee 发表于 2019-2-15 16:47
This is the verbose output from 0.9.8 toolkit.

--> config model

你好，请问你的问题解决了吗？我现在遇到了一样的问题

作者: LSC 时间: 2019-12-19 10:40

chuyee 发表于 2019-2-15 16:47
This is the verbose output from 0.9.8 toolkit.

--> config model

请问你现在解决了吗？

欢迎光临 Toybrick (https://t.rock-chips.com/)