ncnn笔记_cnn代码

　//construct net
　　 ncnn::Net yolov3;
    yolov3.load_param("mobilenetv2_yolov3.param");
    yolov3.load_model("mobilenetv2_yolov3.bin");
  //read in img
    ncnn::Mat in = ncnn::Mat::from_pixels_resize(bgr.data, ncnn::Mat::PIXEL_BGR, bgr.cols, bgr.rows, target_size, target_size);
  //substract mean
    const float mean_vals[3] = {127.5f, 127.5f, 127.5f};
    const float norm_vals[3] = {0.007843f, 0.007843f, 0.007843f};
    in.substract_mean_normalize(mean_vals, norm_vals);
  //construct create_extractor
    ncnn::Extractor ex = yolov3.create_extractor(); 
    ex.set_num_threads(4); 
  //provide input
    ex.input("data", in); 
  //get output
    ncnn::Mat out; ex.extract("detection_out", out);

7767517 15 15 Input data 0 1 data 0=227 1=227 2=3 Convolution conv1 1 1 data conv1_relu1 0=96 1=11 3=4 5=1 6=34848 9=1 LRN norm1 1 1 conv1_relu1 norm1 2=0.000100 Pooling pool1 1 1 norm1 pool1 1=3 2=2 ConvolutionDepthWise conv2 1 1 pool1 conv2_relu2 0=256 1=5 4=2 5=1 6=307200 7=2 9=1 LRN norm2 1 1 conv2_relu2 norm2 2=0.000100 Pooling pool2 1 1 norm2 pool2 1=3 2=2 Convolution conv3 1 1 pool2 conv3_relu3 0=384 1=3 4=1 5=1 6=884736 9=1 ConvolutionDepthWise conv4 1 1 conv3_relu3 conv4_relu4 0=384 1=3 4=1 5=1 6=663552 7=2 9=1 ConvolutionDepthWise conv5 1 1 conv4_relu4 conv5_relu5 0=256 1=3 4=1 5=1 6=442368 7=2 9=1 Pooling pool5 1 1 conv5_relu5 pool5 1=3 2=2 InnerProduct fc6 1 1 pool5 fc6_drop6 0=4096 1=1 2=37748736 9=1 InnerProduct fc7 1 1 fc6_drop6 fc7_drop7 0=4096 1=1 2=16777216 9=1 InnerProduct fc8 1 1 fc7_drop7 fc8 0=1000 1=1 2=4096000 Softmax prob 1 1 fc8 output

int Net::forward_layer(int layer_index, std::vector<Mat>& blob_mats, Option& opt) const { const Layer* layer = layers[layer_index]; if (layer->one_blob_only) { // load bottom blob int bottom_blob_index = layer->bottoms[0]; int top_blob_index = layer->tops[0]; if (blob_mats[bottom_blob_index].dims == 0) { int ret = forward_layer(blobs[bottom_blob_index].producer, blob_mats, opt); if (ret != 0) return ret; } Mat bottom_blob = blob_mats[bottom_blob_index]; // forward if (opt.lightmode && layer->support_inplace) { Mat& bottom_top_blob = bottom_blob; int ret = layer->forward_inplace(bottom_top_blob, opt); if (ret != 0) return ret; // store top blob blob_mats[top_blob_index] = bottom_top_blob; }
　　　}

public:// register custom layer by layer type name // return 0 if success int register_custom_layer(const char* type, layer_creator_func creator); // register custom layer by layer type // return 0 if success int register_custom_layer(int index, layer_creator_func creator); // load network structure from plain param file // return 0 if success int load_param(FILE* fp);protected:

int forward_layer(int layer_index, std::vector<Mat>& blob_mats, Option& opt) const; protected: std::vector<Blob> blobs; std::vector<Layer*> layers; std::vector<layer_registry_entry> custom_layer_registry;

 Layer* layer = create_layer(layer_type); if (!layer) { layer = create_custom_layer(layer_type); } 

//layer.h struct layer_registry_entry { // layer type name const char* name; // layer factory entry  layer_creator_func creator; }; //layer.cpp static const layer_registry_entry layer_registry[] = { #include "layer_registry.h" };

class AbsVal_final : virtual public AbsVal { public: virtual int create_pipeline(const Option& opt) { { int ret = AbsVal::create_pipeline(opt); if (ret) return ret; } return 0; } virtual int destroy_pipeline(const Option& opt) { { int ret = AbsVal::destroy_pipeline(opt); if (ret) return ret; } return 0; } }; DEFINE_LAYER_CREATOR(AbsVal_final) //layer.h #define DEFINE_LAYER_CREATOR(name) \ ::ncnn::Layer* name##_layer_creator() { return new name; }

public: // load layer specific parameter from parsed dict // return 0 if success virtual int load_param(const ParamDict& pd); // load layer specific weight data from model binary // return 0 if success virtual int load_model(const ModelBin& mb); public: // implement inference // return 0 if success virtual int forward(const std::vector<Mat>& bottom_blobs, std::vector<Mat>& top_blobs, const Option& opt = Option()) const; virtual int forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt = Option()) const; // implement inplace inference // return 0 if success virtual int forward_inplace(std::vector<Mat>& bottom_top_blobs, const Option& opt = Option()) const; virtual int forward_inplace(Mat& bottom_top_blob, const Option& opt = Option()) const; public: // layer type index int typeindex; // layer type name std::string type; // layer name std::string name; // blob index which this layer needs as input std::vector<int> bottoms; // blob index which this layer produces as output std::vector<int> tops;