Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.
6539 Discussions

Openvino + ICNet + FPGA Arria 10 GX

jambhale__ranjit
Beginner
‎01-13-2019 09:24 PM
1,508 Views

I am using Arria® 10 GX FPGA Development Kit as accelerator. I have done the setup ready as per instruction given on below link.
https://software.intel.com/en-us/articles/OpenVINO-Install-Linux-FPGA#program%20aria%2010%20gx
I am able to run sample application successfully on FPGA as well.

I do have pre-trained tensorflow model ICNet.pb file. I want to run segmentation application using ICNet network.
Now I was tring to convert ICNet.pb tensorflow network to  ICNet.xml and ICNet.bin file using mo_tf.py tool
But i am getting below mentioned error.


python3 /home/datadrive1/intel/computer_vision_sdk_fpga_2018.4.420/deployment_tools/model_optimizer/mo_tf.py --input_model /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb --data_type FP32 --output_dir .
Model Optimizer arguments:
Common parameters:
    - Path to the Input Model:     /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb
    - Path for generated IR:     /home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/.
    - IR output name:     ICNet-model
    - Log level:     ERROR
    - Batch:     Not specified, inherited from the model
    - Input layers:     Not specified, inherited from the model
    - Output layers:     Not specified, inherited from the model
    - Input shapes:     Not specified, inherited from the model
    - Mean values:     Not specified
    - Scale values:     Not specified
    - Scale factor:     Not specified
    - Precision of IR:     FP32
    - Enable fusing:     True
    - Enable grouped convolutions fusing:     True
    - Move mean values to preprocess section:     False
    - Reverse input channels:     False
TensorFlow specific parameters:
    - Input model in text protobuf format:     False
    - Offload unsupported operations:     False
    - Path to model dump for TensorBoard:     None
    - List of shared libraries with TensorFlow custom layers implementation:     None
    - Update the configuration file with input/output node names:     None
    - Use configuration file used to generate the model with Object Detection API:     None
    - Operations to offload:     None
    - Patterns to offload:     None
    - Use the config file:     None
Model Optimizer version:     1.4.292.6ef7232d
[ ERROR ]  Shape is not defined for output 0 of "split".
[ ERROR ]  Shape is not defined for output 1 of "split".
[ ERROR ]  Shape is not defined for output 2 of "split".
[ ERROR ]  Cannot infer shapes or values for node "split".
[ ERROR ]  Not all output shapes were inferred or fully defined for node "split".
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #40.
[ ERROR ]  
[ ERROR ]  It can happen due to bug in custom shape infer function <function tf_split_infer at 0x7f82df8c1c80>.
[ ERROR ]  Or because the node inputs have incorrect values/shapes.
[ ERROR ]  Or because input shapes are incorrect (embedded to the model or passed via --input_shape).
[ ERROR ]  Run Model Optimizer with --log_level=DEBUG for more information.
[ ERROR ]  Stopped shape/value propagation at "split" node.
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #38.
Thanks & regards
Ranjit Jambhale

0 Kudos

Link Copied

5 Replies
jambhale__ranjit
Beginner
‎01-13-2019 09:28 PM
1,508 Views

Please find the error in detailed

Command :    python3 /home/datadrive1/intel/computer_vision_sdk_fpga_2018.4.420/deployment_tools/model_optimizer/mo_tf.py --input_model /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb --data_type FP32 --output_dir .
Model Optimizer arguments:
Common parameters:
    - Path to the Input Model:     /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb
    - Path for generated IR:     /home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/.
    - IR output name:     ICNet-model
    - Log level:     ERROR
    - Batch:     Not specified, inherited from the model
    - Input layers:     Not specified, inherited from the model
    - Output layers:     Not specified, inherited from the model
    - Input shapes:     Not specified, inherited from the model
    - Mean values:     Not specified
    - Scale values:     Not specified
    - Scale factor:     Not specified
    - Precision of IR:     FP32
    - Enable fusing:     True
    - Enable grouped convolutions fusing:     True
    - Move mean values to preprocess section:     False
    - Reverse input channels:     False
TensorFlow specific parameters:
    - Input model in text protobuf format:     False
    - Offload unsupported operations:     False
    - Path to model dump for TensorBoard:     None
    - List of shared libraries with TensorFlow custom layers implementation:     None
    - Update the configuration file with input/output node names:     None
    - Use configuration file used to generate the model with Object Detection API:     None
    - Operations to offload:     None
    - Patterns to offload:     None
    - Use the config file:     None
Model Optimizer version:     1.4.292.6ef7232d
[ ERROR ]  Shape is not defined for output 0 of "split".
[ ERROR ]  Shape is not defined for output 1 of "split".
[ ERROR ]  Shape is not defined for output 2 of "split".
[ ERROR ]  Cannot infer shapes or values for node "split".
[ ERROR ]  Not all output shapes were inferred or fully defined for node "split".
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #40.
[ ERROR ]  
[ ERROR ]  It can happen due to bug in custom shape infer function <function tf_split_infer at 0x7f82df8c1c80>.
[ ERROR ]  Or because the node inputs have incorrect values/shapes.
[ ERROR ]  Or because input shapes are incorrect (embedded to the model or passed via --input_shape).
[ ERROR ]  Run Model Optimizer with --log_level=DEBUG for more information.
[ ERROR ]  Stopped shape/value propagation at "split" node.
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #38.

0 Kudos
Copy link
CJian1
Beginner
‎01-14-2019 04:40 PM
1,508 Views

You need to define "--input_shape" if the shape of input is undefined in your tf code (the batch size was probably set to be "None"). And if you have multiple inputs, you should also set the "--input" option as well. Try something like this:

python3 mo.py --input_model /path-to/your-model.caffemodel --input data,rois --input_shape (1,3,227,227),(1,6,1,1) 

0 Kudos
Copy link
jambhale__ranjit
Beginner
‎01-15-2019 09:45 PM
1,508 Views

I have tried the way you suggested but getting error as below

python3 /home/datadrive1/intel/computer_vision_sdk_fpga_2018.4.420/deployment_tools/model_optimizer/mo.py --input_model /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb --data_type FP32 --output_dir . --input split --input_shape "[1,3,227,227]","[1,6,1,1]"
Model Optimizer arguments:
Common parameters:
    - Path to the Input Model:     /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb
    - Path for generated IR:     /home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/.
    - IR output name:     ICNet-model
    - Log level:     ERROR
    - Batch:     Not specified, inherited from the model
    - Input layers:     split
    - Output layers:     Not specified, inherited from the model
    - Input shapes:     [1,3,227,227],[1,6,1,1]
    - Mean values:     Not specified
    - Scale values:     Not specified
    - Scale factor:     Not specified
    - Precision of IR:     FP32
    - Enable fusing:     True
    - Enable grouped convolutions fusing:     True
    - Move mean values to preprocess section:     False
    - Reverse input channels:     False
TensorFlow specific parameters:
    - Input model in text protobuf format:     False
    - Offload unsupported operations:     False
    - Path to model dump for TensorBoard:     None
    - List of shared libraries with TensorFlow custom layers implementation:     None
    - Update the configuration file with input/output node names:     None
    - Use configuration file used to generate the model with Object Detection API:     None
    - Operations to offload:     None
    - Patterns to offload:     None
    - Use the config file:     None
Model Optimizer version:     1.4.292.6ef7232d
[ ERROR ]  Please provide each input layers with an input layer shape.
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #58.

0 Kudos
Copy link
CJian1
Beginner
‎01-16-2019 03:21 AM
1,508 Views

Since you have two input shape, you will also need to specify two inputs after "--input".

0 Kudos
Copy link
jambhale__ranjit
Beginner
‎01-16-2019 09:06 PM
1,509 Views

Before adding --input parameter i was getting error as below :
python3 /home/datadrive1/intel/computer_vision_sdk_fpga_2018.4.420/deployment_tools/model_optimizer/mo.py --input_model /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb --data_type FP32 --output_dir . --input split/split_dim --input_shape "[1,227,227,3]","[1,6,1,1]" --log_level=DEBUG

[ ERROR ]  Shape is not defined for output 0 of "split".
[ ERROR ]  Shape is not defined for output 1 of "split".
[ ERROR ]  Shape is not defined for output 2 of "split".
[ ERROR ]  Cannot infer shapes or values for node "split".
[ ERROR ]  Not all output shapes were inferred or fully defined for node "split".
 For more information please refer to Model Optimizer FAQ (<INSTALL_DIR>/deployment_tools/documentation/docs/MO_FAQ.html), question #40.
[ ERROR ]  
[ ERROR ]  It can happen due to bug in custom shape infer function <function tf_split_infer at 0x7fb36ff52d08>.
[ ERROR ]  Or because the node inputs have incorrect values/shapes.
[ ERROR ]  Or because input shapes are incorrect (embedded to the model or passed via --input_shape).
[ 2019-01-16 17:45:18,363 ] [ DEBUG ] [ infer:198 ]  Node "split" attributes: {'infer': <function tf_split_infer at 0x7fb36ff52d08>, 'name': 'split', 'pb': name: "split"
op: "Split"
input: "split/split_dim"
input: "Placeholder"
attr {
  key: "T"
  value {
    type: DT_FLOAT
  }
}
attr {
  key: "num_split"
  value {
    i: 3
  }
}


After adding --input parameter i was getting error as below :

python3 /home/datadrive1/intel/computer_vision_sdk_fpga_2018.4.420/deployment_tools/model_optimizer/mo.py --input_model /home/datadrive1/prj_IntelFPGA/Deewakar_shared_Files/ICNet-model.pb --data_type FP32 --output_dir . --input split/split_dim,Placeholder --input_shape "[1,227,227,3]","[1,6,1,1]" --log_level=DEBUG

 2019-01-17 10:30:05,994 ] [ DEBUG ] [ register_custom_ops:105 ]  Added a new entry TensorIteratorInput to extractors with custom op class <class 'extensions.ops.TensorIterator_ops.TensorIteratorInput'>.
[ 2019-01-17 10:30:06,106 ] [ DEBUG ] [ extractor:829 ]  Sink: Reshape_1/sink_port_0 for node Reshape_1
[ 2019-01-17 10:30:06,107 ] [ DEBUG ] [ extractor:830 ]  {'is_output': True, 'data_type': None, 'infer': None, 'value': None, 'type': 'OpOutput', 'precision': 'FP32', 'dim_attrs': ['batch_dims', 'axis', 'channel_dims', 'spatial_dims'], 'kind': 'op', 'op': 'OpOutput', 'IE': [('layer', [('id', <function Op.substitute_ie_attrs.<locals>.<lambda> at 0x7f10ccad32f0>), 'name', 'precision', 'type'], [('data', [], []), '@ports', '@consts'])], 'shape_attrs': ['shape', 'window', 'pad', 'output_shape', 'stride'], 'name': 'Reshape_1/sink_port_0'}
[ 2019-01-17 10:30:06,107 ] [ DEBUG ] [ extractor:831 ]  Add edge from Reshape_1 to Reshape_1/sink_port_0
[ ERROR ]  -------------------------------------------------
[ ERROR ]  ----------------- INTERNAL ERROR ----------------
[ ERROR ]  Unexpected exception happened.
[ ERROR ]  Please contact Model Optimizer developers and forward the following information:
[ ERROR ]  
[ ERROR ]  Traceback (most recent call last):
  File "/home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/mo/main.py", line 325, in main
    return driver(argv)
  File "/home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/mo/main.py", line 267, in driver
    mean_scale_values=mean_scale)
  File "/home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/mo/pipeline/tf.py", line 176, in tf2nx
    input_op_nodes = add_input_ops(graph, packed_user_shapes, True)
  File "/home/datadrive1/intel/computer_vision_sdk_2018.4.420/deployment_tools/model_optimizer/mo/front/extractor.py", line 993, in add_input_ops
    assert n_inputs == 1
AssertionError

[ ERROR ]  ---------------- END OF BUG REPORT --------------
[ ERROR ]  -------------------------------------------------

Queries

1. Does --input parameter am i using are correct ?
2. Currently i want to convert tensorflow ICNet.pb model to .xml and .bin using mo.py .
   Does --input_shape parameter am i using are correct ?


Thanks & regards

Ranjit

0 Kudos
Copy link
Reply

Community support is provided Monday to Friday. Other contact methods are available here.

Intel does not verify all solutions, including but not limited to any file transfers that may appear in this community. Accordingly, Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.

For more complete information about compiler optimizations, see our Optimization Notice.