Thank you for reaching out! Now that you have your TensorFlow model, you will need to convert it to an Intel Movidius Graph file that can be used with the Neural Compute Stick.
The Neural Compute SDK includes the mvNCCompile that is used to compile the networks into the Graph.
Take a look at the Tools provided with the NCSDK.
Once you have a graph file, you will need to write a C or Python program to load the graph into the Neural Compute Stick. I would start by looking at the examples included in the NCSDK and NCAPPZOO.
Hope this helps!
Thanks! It works! :) But I meet an error. :'(
When I command
mvNCCompile frozen_graph_slim_87%.pb -s 12 -in=input -on=InceptionV2/Predictions/Reshape_1 -is 224 224 -o inception_v2.graph. Errors have arisen.
Caused by op 'InceptionV2/InceptionV2/Conv2d_1a_7x7/separable_conv2d/depthwise', defined at:
File "/usr/local/bin/mvNCCompile", line 118, in
create_graph(args.network, args.inputnode, args.outputnode, args.outfile, args.nshaves, args.inputsize, args.weights)
File "/usr/local/bin/mvNCCompile", line 104, in create_graph
net = parse_tensor(args, myriad_config)
File "/usr/local/bin/ncsdk/Controllers/TensorFlowParser.py", line 211, in parse_tensor
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/importer.py", line 313, in import_graph_def
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/ops.py", line 2956, in create_op
File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/ops.py", line 1470, in init
self._traceback = self._graph._extract_stack() # pylint: disable=protected-access
InvalidArgumentError (see above for traceback): NodeDef mentions attr 'dilations' not in Op output:T; attr=T:type,allowed=[DT_FLOAT, DT_DOUBLE]; attr=strides:list(int); attr=padding:string,allowed=["SAME", "VALID"]; attr=data_format:string,default="NHWC",allowed=["NHWC", "NCHW"]>; NodeDef: InceptionV2/InceptionV2/Conv2d_1a_7x7/separable_conv2d/depthwise = DepthwiseConv2dNative[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="SAME", strides=[1, 2, 2, 1], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_input_0_0, InceptionV2/Conv2d_1a_7x7/depthwise_weights). (Check whether your GraphDef-interpreting binary is up to date with your GraphDef-generating binary.).
[[Node: InceptionV2/InceptionV2/Conv2d_1a_7x7/separable_conv2d/depthwise = DepthwiseConv2dNative[T=DT_FLOAT, data_format="NHWC", dilations=[1, 1, 1, 1], padding="SAME", strides=[1, 2, 2, 1], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_input_0_0, InceptionV2/Conv2d_1a_7x7/depthwise_weights)]]
I use bazel to check OutputNode/InputNode of my model. Information follows:
Found 1 possible inputs: (name=input, type=float(1), shape=[1,224,224,3])
No variables spotted.
Found 1 possible outputs: (name=InceptionV2/Predictions/Reshape_1, op=Reshape)
Found 10187114 (10.19M) const parameters, 0 (0) variable parameters, and 0 control_edges
Op types used: 357 Const, 278 Identity, 70 Conv2D, 69 Relu, 68 FusedBatchNorm, 10 ConcatV2, 8 AvgPool, 5 MaxPool, 2 BiasAdd, 2 Reshape, 1 DepthwiseConv2dNative, 1 Placeholder, 1 Softmax, 1 Squeeze
To use with tensorflow/tools/benchmark:benchmark_model try these arguments:
bazel run tensorflow/tools/benchmark:benchmark_model -- --graph=/home/xuyifang/Desktop/ncs/frozen_graph_slim_87%.pb --show_flops --input_layer=input --input_layer_type=float --input_layer_shape=1,224,224,3 --output_layer=InceptionV2/Predictions/Reshape_1
Can you help me? Thank u verrrrrrrrrrrrry much!
Could you share the model you are trying to compile? I would like to try this myself.
Which Neural Compute SDK version are you using? Are you using the Neural Compute Stick 1 or 2?
I used tensorflow1.5 to solve the last problem. But I face a new one. All predicted picture probabilities are nan. I don't know why, because it's still correct on my computer.
This is result.
prediction 0(probability nan) is 3 label index is: 3
prediction 1(probability nan) is 2 label index is: 2
prediction 2(probability nan) is 1 label index is: 1
prediction 3(probability nan) is 0 label index is: 0
Can u help me? That a looooot!!!!!
Can you share the model and steps to reproduce?
The following thread seems to be similar to the issue you are seeing.