yolo源码解析(1):代码逻辑

一. 整体代码逻辑

yolo中源码分为三个部分,\example,\include,以及\src文件夹下都有源代码存在.

结构如下所示

├── examples
│   ├── darknet.c(主程序)
│   │── xxx1.c
│   └── xxx2.c
│
├── include
│   ├── darknet.h
│ 
│ 
├── Makefile
│ 
│ 
└── src
    ├── yyy1.c
    ├── yyy2.h
    └──......

 

\include文件夹中没有.h头文件, 里边的内容算作一个整体, 都是darknet.c中的一部分, 每个文件的内容共darknet.c调用, 除了darknet.c外, \include文件夹中的文件不存在互相调用, 各自完成不同的功能,如检测视频, 检测图片, 检测艺术品等, 通过darknet.c中的if条件进行选择调用. 因为这部分算作一个整体, 所以共用darknet.h这个头文件. 如果\include需要用到\src中的函数, 则在darknet.h中进行声明

在\src文件夹中, 每个c文件都对应一个同名的.h头文件; main函数存在于\example文件夹下的darknet.c文件中.

\include文件夹下的darknet.h的作用是联系\example与\src两部分, 在这两部分中都需要用的函数则在darknet.h中进行声明, 例如\example中有xxx1.c, \src中有yyy1.c及yyy1.h, xxx1.c与yyy1.c中都需要用到func()这个函数, 那么func()的声明需要放在darknet.h中, 然后在xxx1.c与yyy1.h分别引入头文件#include "darknet.h"

而如果\example\darknet.c中需要调用\example\xxx1.c中的函数, 则需要在\example\darknet.c加extern字段

 

多文件的实现方式(头文件的使用)

在本项目中, \includes\darknet.h是\examples中文件的头文件, 而在\includes\darknet.h中, 又对部分函数(例如 void forward_network(network *net); )进行了声明, 但是 forward_network 函数的定义是在\src\network.c中, 因为定义是在\src中, 所以定义时\src中的文件需要引入darknet.h这个头文件; 由此, \examples中的文件便可通过darknet.h中的声明调用\src中的函数了

举例

对于 ./darknet detect cfg/yolov3.cfg yolov3.weights data/dog.jpg 这条命令,首先传送到darknet.c文件, 然后darknet.c文件检测到含有detect字符, 所以进入if语句. 使用\src\utils.c中的find_char_arg函数来获取输出文件名等信息, 然后调用detector.c文件中的test_detector函数, 该函数负责检测并进行输出.

 

二. main函数

 唉唉唉

 

 

三. makefile文件

入门见<并行程序设计(第四版)>

以yolo源码中的makefile文件为例

GPU=0
CUDNN=0
OPENCV=0
OPENMP=0
DEBUG=0

ARCH= -gencode arch=compute_30,code=sm_30 \
      -gencode arch=compute_35,code=sm_35 \
      -gencode arch=compute_50,code=[sm_50,compute_50] \
      -gencode arch=compute_52,code=[sm_52,compute_52]
#      -gencode arch=compute_20,code=[sm_20,sm_21] \ This one is deprecated?

# This is what I use, uncomment if you know your arch and want to specify
# ARCH= -gencode arch=compute_52,code=compute_52

VPATH=./src/:./examples
# VTATH用来告诉make,源文件的路径, 参考https://blog.csdn.net/mcgrady_tracy/article/details/27240139
SLIB=libdarknet.so
ALIB=libdarknet.a
EXEC=darknet
OBJDIR=./obj/

CC=gcc
NVCC=nvcc 
AR=ar
ARFLAGS=rcs
OPTS=-Ofast
LDFLAGS= -lm -pthread       #gcc等编译器会用到的一些优化参数，也可以在里面指定库文件的位置
COMMON= -Iinclude/ -Isrc/
CFLAGS=-Wall -Wno-unused-result -Wno-unknown-pragmas -Wfatal-errors -fPIC   #指定头文件（.h文件）的路径，如：CFLAGS=-I/usr/include -I/path/include。

ifeq ($(OPENMP), 1) 
CFLAGS+= -fopenmp
endif

ifeq ($(DEBUG), 1) 
OPTS=-O0 -g
endif

CFLAGS+=$(OPTS)

ifeq ($(OPENCV), 1) 
COMMON+= -DOPENCV
CFLAGS+= -DOPENCV
LDFLAGS+= `pkg-config --libs opencv` 
COMMON+= `pkg-config --cflags opencv` 
endif

ifeq ($(GPU), 1) 
COMMON+= -DGPU -I/usr/local/cuda/include/
CFLAGS+= -DGPU
LDFLAGS+= -L/usr/local/cuda/lib64 -lcuda -lcudart -lcublas -lcurand
endif

ifeq ($(CUDNN), 1) 
COMMON+= -DCUDNN 
CFLAGS+= -DCUDNN
LDFLAGS+= -lcudnn
endif

OBJ=gemm.o utils.o cuda.o deconvolutional_layer.o convolutional_layer.o list.o image.o activations.o im2col.o col2im.o blas.o crop_layer.o dropout_layer.o maxpool_layer.o softmax_layer.o data.o matrix.o network.o connected_layer.o cost_layer.o parser.o option_list.o detection_layer.o route_layer.o upsample_layer.o box.o normalization_layer.o avgpool_layer.o layer.o local_layer.o shortcut_layer.o logistic_layer.o activation_layer.o rnn_layer.o gru_layer.o crnn_layer.o demo.o batchnorm_layer.o region_layer.o reorg_layer.o tree.o  lstm_layer.o l2norm_layer.o yolo_layer.o
EXECOBJA=my_test.o  captcha.o lsd.o super.o art.o tag.o cifar.o go.o rnn.o segmenter.o regressor.o classifier.o coco.o yolo.o detector.o nightmare.o darknet.o
ifeq ($(GPU), 1) 
LDFLAGS+= -lstdc++ 
OBJ+=convolutional_kernels.o deconvolutional_kernels.o activation_kernels.o im2col_kernels.o col2im_kernels.o blas_kernels.o crop_layer_kernels.o dropout_layer_kernels.o maxpool_layer_kernels.o avgpool_layer_kernels.o
endif

EXECOBJ = $(addprefix $(OBJDIR), $(EXECOBJA))   #加前缀函数: $(addprefix <prefix>,<names...>),OBJDIR是obj存放的地址
OBJS = $(addprefix $(OBJDIR), $(OBJ))
DEPS = $(wildcard src/*.h) Makefile include/darknet.h

#all: obj backup results $(SLIB) $(ALIB) $(EXEC)
all: obj  results $(SLIB) $(ALIB) $(EXEC)


$(EXEC): $(EXECOBJ) $(ALIB)
    $(CC) $(COMMON) $(CFLAGS) $^ -o $@ $(LDFLAGS) $(ALIB)

$(ALIB): $(OBJS)
    $(AR) $(ARFLAGS) $@ $^

$(SLIB): $(OBJS)
    $(CC) $(CFLAGS) -shared $^ -o $@ $(LDFLAGS)

$(OBJDIR)%.o: %.c $(DEPS)
    $(CC) $(COMMON) $(CFLAGS) -c $< -o $@

$(OBJDIR)%.o: %.cu $(DEPS)
    $(NVCC) $(ARCH) $(COMMON) --compiler-options "$(CFLAGS)" -c $< -o $@

obj:
    mkdir -p obj
backup:
    mkdir -p backup
results:
    mkdir -p results

.PHONY: clean

clean:
    rm -rf $(OBJS) $(SLIB) $(ALIB) $(EXEC) $(EXECOBJ) $(OBJDIR)/*

View Code

关于vpath,参考https://blog.csdn.net/mcgrady_tracy/article/details/27240139

(1)修改代码的第一次尝试

在\examples文件夹下新建my_test.c文件, 内容如下

#include "darknet.h"

void output_to_file()
{
    FILE *fp;
    fp=fopen("output.txt","w");
    fprintf(fp,"adfsss");
    printf("test\n");
    fclose(fp);
}

View Code

在darknet.c中进行调用, 如下

#include "darknet.h"

#include <time.h>
#include <stdlib.h>
#include <stdio.h>
//

extern void predict_classifier(char *datacfg, char *cfgfile, char *weightfile, char *filename, int top);     // 在\examples\classifier.c中
extern void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh, float hier_thresh, char *outfile, int fullscreen);     // 在\examples\detector.c中
extern void run_yolo(int argc, char **argv);     // 在\examples\yolo.c中
extern void run_detector(int argc, char **argv);     // 在\examples\detector.c中
extern void run_coco(int argc, char **argv);         // 在\examples\coco.c中
extern void run_captcha(int argc, char **argv);      // 在\examples\captcha.c中
extern void run_nightmare(int argc, char **argv);        // 在\examples\nightmare.c中
extern void run_classifier(int argc, char **argv);       // 在\examples\classifier.c中
extern void run_regressor(int argc, char **argv);        // 在\examples\regressor.c中
extern void run_segmenter(int argc, char **argv);        // 在\examples\segmenter.c中
extern void run_char_rnn(int argc, char **argv);         // 在\examples\rnn.c中
extern void run_tag(int argc, char **argv);              // 在\examples\tag.c中
extern void run_cifar(int argc, char **argv);            // 在\examples\fun_cifar.c中
extern void run_go(int argc, char **argv);               // 在\examples\go.c中
extern void run_art(int argc, char **argv);              // 在\examples\art.c中
extern void run_super(int argc, char **argv);            // 在\examples\super.c中
extern void run_lsd(int argc, char **argv);              // 在\examples\nightmare.c中
extern void output_to_file();              // 在\examples\my_test.c中

void average(int argc, char *argv[])
{
    char *cfgfile = argv[2];
    char *outfile = argv[3];
    gpu_index = -1;
    network *net = parse_network_cfg(cfgfile);
    network *sum = parse_network_cfg(cfgfile);

    char *weightfile = argv[4];   
    load_weights(sum, weightfile);

    int i, j;
    int n = argc - 5;
    for(i = 0; i < n; ++i){
        weightfile = argv[i+5];   
        load_weights(net, weightfile);
        for(j = 0; j < net->n; ++j){
            layer l = net->layers[j];
            layer out = sum->layers[j];
            if(l.type == CONVOLUTIONAL){
                int num = l.n*l.c*l.size*l.size;
                axpy_cpu(l.n, 1, l.biases, 1, out.biases, 1);
                axpy_cpu(num, 1, l.weights, 1, out.weights, 1);
                if(l.batch_normalize){
                    axpy_cpu(l.n, 1, l.scales, 1, out.scales, 1);
                    axpy_cpu(l.n, 1, l.rolling_mean, 1, out.rolling_mean, 1);
                    axpy_cpu(l.n, 1, l.rolling_variance, 1, out.rolling_variance, 1);
                }
            }
            if(l.type == CONNECTED){
                axpy_cpu(l.outputs, 1, l.biases, 1, out.biases, 1);
                axpy_cpu(l.outputs*l.inputs, 1, l.weights, 1, out.weights, 1);
            }
        }
    }
    n = n+1;
    for(j = 0; j < net->n; ++j){
        layer l = sum->layers[j];
        if(l.type == CONVOLUTIONAL){
            int num = l.n*l.c*l.size*l.size;
            scal_cpu(l.n, 1./n, l.biases, 1);
            scal_cpu(num, 1./n, l.weights, 1);
                if(l.batch_normalize){
                    scal_cpu(l.n, 1./n, l.scales, 1);
                    scal_cpu(l.n, 1./n, l.rolling_mean, 1);
                    scal_cpu(l.n, 1./n, l.rolling_variance, 1);
                }
        }
        if(l.type == CONNECTED){
            scal_cpu(l.outputs, 1./n, l.biases, 1);
            scal_cpu(l.outputs*l.inputs, 1./n, l.weights, 1);
        }
    }
    save_weights(sum, outfile);
}

long numops(network *net)
{
    int i;
    long ops = 0;
    for(i = 0; i < net->n; ++i){
        layer l = net->layers[i];
        if(l.type == CONVOLUTIONAL){
            ops += 2l * l.n * l.size*l.size*l.c/l.groups * l.out_h*l.out_w;
        } else if(l.type == CONNECTED){
            ops += 2l * l.inputs * l.outputs;
        } else if (l.type == RNN){
            ops += 2l * l.input_layer->inputs * l.input_layer->outputs;
            ops += 2l * l.self_layer->inputs * l.self_layer->outputs;
            ops += 2l * l.output_layer->inputs * l.output_layer->outputs;
        } else if (l.type == GRU){
            ops += 2l * l.uz->inputs * l.uz->outputs;
            ops += 2l * l.uh->inputs * l.uh->outputs;
            ops += 2l * l.ur->inputs * l.ur->outputs;
            ops += 2l * l.wz->inputs * l.wz->outputs;
            ops += 2l * l.wh->inputs * l.wh->outputs;
            ops += 2l * l.wr->inputs * l.wr->outputs;
        } else if (l.type == LSTM){
            ops += 2l * l.uf->inputs * l.uf->outputs;
            ops += 2l * l.ui->inputs * l.ui->outputs;
            ops += 2l * l.ug->inputs * l.ug->outputs;
            ops += 2l * l.uo->inputs * l.uo->outputs;
            ops += 2l * l.wf->inputs * l.wf->outputs;
            ops += 2l * l.wi->inputs * l.wi->outputs;
            ops += 2l * l.wg->inputs * l.wg->outputs;
            ops += 2l * l.wo->inputs * l.wo->outputs;
        }
    }
    return ops;
}

void speed(char *cfgfile, int tics)
{
    if (tics == 0) tics = 1000;
    network *net = parse_network_cfg(cfgfile);
    set_batch_network(net, 1);
    int i;
    double time=what_time_is_it_now();
    image im = make_image(net->w, net->h, net->c*net->batch);
    for(i = 0; i < tics; ++i){
        network_predict(net, im.data);
    }
    double t = what_time_is_it_now() - time;
    long ops = numops(net);
    printf("\n%d evals, %f Seconds\n", tics, t);
    printf("Floating Point Operations: %.2f Bn\n", (float)ops/1000000000.);
    printf("FLOPS: %.2f Bn\n", (float)ops/1000000000.*tics/t);
    printf("Speed: %f sec/eval\n", t/tics);
    printf("Speed: %f Hz\n", tics/t);
}

void operations(char *cfgfile)
{
    gpu_index = -1;
    network *net = parse_network_cfg(cfgfile);
    long ops = numops(net);
    printf("Floating Point Operations: %ld\n", ops);
    printf("Floating Point Operations: %.2f Bn\n", (float)ops/1000000000.);
}

void oneoff(char *cfgfile, char *weightfile, char *outfile)
{
    gpu_index = -1;
    network *net = parse_network_cfg(cfgfile);
    int oldn = net->layers[net->n - 2].n;
    int c = net->layers[net->n - 2].c;
    scal_cpu(oldn*c, .1, net->layers[net->n - 2].weights, 1);
    scal_cpu(oldn, 0, net->layers[net->n - 2].biases, 1);
    net->layers[net->n - 2].n = 11921;
    net->layers[net->n - 2].biases += 5;
    net->layers[net->n - 2].weights += 5*c;
    if(weightfile){
        load_weights(net, weightfile);
    }
    net->layers[net->n - 2].biases -= 5;
    net->layers[net->n - 2].weights -= 5*c;
    net->layers[net->n - 2].n = oldn;
    printf("%d\n", oldn);
    layer l = net->layers[net->n - 2];
    copy_cpu(l.n/3, l.biases, 1, l.biases +   l.n/3, 1);
    copy_cpu(l.n/3, l.biases, 1, l.biases + 2*l.n/3, 1);
    copy_cpu(l.n/3