本文主要是介绍intel Movidius Neural Network Stick各种网络速度测试，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
make examples
| 网络名 | time(ms) |
|---|---|
| caffe_alexnet | 88.04 |
| caffe_googlelenet | 93.65 |
| caffe_squeezenet | 46.70 |
| tf_inception_v3 | 561.82 |
| tf_inception_v1 | 88.57 |
ncappzoo
| 网络名 | time(ms) |
|---|---|
| caffe_tinyYolo | 117.41 |
| caffe_SSD_mobilenet | 80.15 |
| tf_Mobilenet_v1_1.0_224 | 39.97 |
| tf_Mobilenet_v1_0.5_224 | 17.97 |
tensorflow ssd_mobilenet_v1_coco_2017_11_17移植到NCS计算棒遇到的问题
[Error 5] Toolkit Error: Stage Details Not Supported: Cast
该问题是因为模型中使用了NCS不支持的操作引起的，从mvNCCompile源码中找到支持的tensorflow操作类型总共包括一下几种：
[‘Add’, ‘Assign’, ‘AvgPool’, ‘BiasAdd’, ‘Concat’, ‘ConcatV2’, ‘Const’, ‘Conv2D’, ‘Conv2DBackpropInput’, ‘DepthwiseConv2dNative’, ‘Div’, ‘ExpandDims’, ‘ExtractImagePatches’, ‘Fill’, ‘Floor’, ‘FusedBatchNorm’, ‘Identity’, ‘L2Loss’, ‘LRN’, ‘MatMul’, ‘MaxPool’, ‘Maximum’, ‘Mean’, ‘Mul’, ‘NoOp’, ‘Pack’, ‘Pad’, ‘Placeholder’, ‘PlaceholderWithDefault’, ‘Prod’, ‘RandomStandardNormal’, ‘RandomUniform’, ‘RealDiv’, ‘Relu’, ‘Relu6’, ‘Reshape’, ‘Rsqrt’, ‘Shape’, ‘Sigmoid’, ‘Slice’, ‘Softmax’, ‘Square’, ‘Squeeze’, ‘StridedSlice’, ‘Sub’, ‘Sum’, ‘Tanh’, ‘TruncatedNormal’, ‘Variable’, ‘VariableV2’]
这篇关于intel Movidius Neural Network Stick各种网络速度测试的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
