-
开发环境:
ubuntu 20.04
vscode
rtx2080ti
怎么配置看之前的文章 -
darknet_ros环境
2.1 下载
mkdir -p darknet_ros_yolov3/src
cd darknet_ros_yolov3/src
git clone --recursive git@github.com:leggedrobotics/darknet_ros.git
注意: git clone 这里要加–recursive，不然后面会报错找不到啥啥啥
2.2 编译darknet的Makefile,主要是支持gpu，就想用cpu就直接编译，我的Makefile文件如下
GPU=1
CUDNN=1
OPENCV=1
OPENMP=1
DEBUG=0
ARCH= -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_70,code=[sm_70,compute_70] \
-gencode arch=compute_75,code=[sm_75,compute_75]\
-gencode arch=compute_86,code=[sm_86,compute_86]
# 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
SLIB=libdarknet.so
ALIB=libdarknet.a
EXEC=darknet
OBJDIR=./obj/
CC=gcc
CPP=g++
NVCC=nvcc
AR=ar
ARFLAGS=rcs
OPTS=-Ofast
LDFLAGS= -lm -pthread
COMMON= -Iinclude/ -Isrc/
CFLAGS=-Wall -Wno-unused-result -Wno-unknown-pragmas -Wfatal-errors -fPIC
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 opencv4` -lstdc++
COMMON+= `pkg-config --cflags opencv4`
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 iseg_layer.o image_opencv.o
EXECOBJA=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 instance-segmenter.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))
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: %.cpp $(DEPS)
$(CPP) $(COMMON) $(CFLAGS) -c $< -o $@
$(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)/*
查看显卡匹配的arch和gencode:
Matching CUDA arch and CUDA gencode for various NVIDIA architectures
这里要去掉30那个，说是过期了，我注释掉重新编译可以成功。
接下来编译
cd darknet_ros_yolov3/darknet
make
2.3 编译darknet_ros
catkin_make -DCMAKE_BUILD_TYPE=Release -DCMAKE_EXPORT_COMPILE_COMMANDS=Yes
我后来又编译了debug版本，直接type换成Debug也是过去了。
期间需要下载cfg和weights，可以下好了放对应对方，再重新编译一遍就可以了。
2022年2月10更新：
这里有个报错:
如何解决pjreddie版darknet不能使用cudnn8编译的问题
参考这个博客，主要思路是加上支持的处理:
如何解决pjreddie版darknet不能使用cudnn8编译的问题
我的src/convolutional_layer.c文件如下
#include "convolutional_layer.h"
#include "utils.h"
#include "batchnorm_layer.h"
#include "im2col.h"
#include "col2im.h"
#include "blas.h"
#include "gemm.h"
#include <stdio.h>
#include <time.h>
#define PRINT_CUDNN_ALGO 0
#define MEMORY_LIMIT 2000000000
#ifdef AI2
#include "xnor_layer.h"
#endif
void swap_binary(convolutional_layer *l)
{
float *swap = l->weights;
l->weights = l->binary_weights;
l->binary_weights = swap;
#ifdef GPU
swap = l->weights_gpu;
l->weights_gpu = l->binary_weights_gpu;
l->binary_weights_gpu = swap;
#endif
}
void binarize_weights(float *weights, int n, int size, float *binary)
{
int i, f;
for(f = 0; f < n; ++f){
float mean = 0;
for(i = 0; i < size; ++i){
mean += fabs(weights[f*size + i]);
}
mean = mean / size;
for(i = 0; i < size; ++i){
binary[f*size + i] = (weights[f*size + i] > 0) ? mean : -mean;
}
}
}
void binarize_cpu(float *input, int n, float *binary)
{
int i;
for(i = 0; i < n; ++i){
binary[i] = (input[i] > 0) ? 1 : -1;
}
}
void binarize_input(float *input, int n, int size, float *binary)
{
int i, s;
for(s = 0; s < size; ++s){
float mean = 0;
for(i = 0; i < n; ++i){
mean += fabs(input[i*size + s]);
}
mean = mean / n;
for(i = 0; i < n; ++i){
binary[i*size + s] = (input[i*size + s] > 0) ? mean : -mean;
}
}
}
int convolutional_out_height(convolutional_layer l)
{
return (l.h + 2*l.pad - l.size) / l.stride + 1;
}
int convolutional_out_width(convolutional_layer l)
{
return (l.w + 2*l.pad - l.size) / l.stride + 1;
}
image get_convolutional_image(convolutional_layer l)
{
return float_to_image(l.out_w,l.out_h,l.out_c,l.output);
}
image get_convolutional_delta(convolutional_layer l)
{
return float_to_image(l.out_w,l.out_h,l.out_c,l.delta);
}
static size_t get_workspace_size(layer l){
#ifdef CUDNN
if(gpu_index >= 0){
size_t most = 0;
size_t s = 0;
cudnnGetConvolutionForwardWorkspaceSize(cudnn_handle(),
l.srcTensorDesc,
l.weightDesc,
l.convDesc,
l.dstTensorDesc,
l.fw_algo,
&s);
if (s > most) most = s;
cudnnGetConvolutionBackwardFilterWorkspaceSize(cudnn_handle(),
l.srcTensorDesc,
l.ddstTensorDesc,
l.convDesc,
l.dweightDesc,
l.bf_algo,
&s);
if (s > most) most = s;
cudnnGetConvolutionBackwardDataWorkspaceSize(cudnn_handle(),
l.weightDesc,
l.ddstTensorDesc,
l.convDesc,
l.dsrcTensorDesc,
l.bd_algo,
&s);
if (s > most) most = s;
return most;
}
#endif
return (size_t)l.out_h*l.out_w*l.size*l.size*l.c/l.groups*sizeof(float);
}
#ifdef GPU
#ifdef CUDNN
void cudnn_convolutional_setup(layer *l)
{
cudnnSetTensor4dDescriptor(l->dsrcTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, l->batch, l->c, l->h, l->w);
cudnnSetTensor4dDescriptor(l->ddstTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, l->batch, l->out_c, l->out_h, l->out_w);
cudnnSetTensor4dDescriptor(l->srcTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, l->batch, l->c, l->h, l->w);
cudnnSetTensor4dDescriptor(l->dstTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, l->batch, l->out_c, l->out_h, l->out_w);
cudnnSetTensor4dDescriptor(l->normTensorDesc, CUDNN_TENSOR_NCHW, CUDNN_DATA_FLOAT, 1, l->out_c, 1, 1);
cudnnSetFilter4dDescriptor(l->dweightDesc, CUDNN_DATA_FLOAT, CUDNN_TENSOR_NCHW, l->n, l->c/l->groups, l->size, l->size);
cudnnSetFilter4dDescriptor(l->weightDesc, CUDNN_DATA_FLOAT, CUDNN_TENSOR_NCHW, l->n, l->c/l->groups, l->size, l->size);
#if CUDNN_MAJOR >= 6
cudnnSetConvolution2dDescriptor(l->convDesc, l->pad, l->pad, l->stride, l->stride, 1, 1, CUDNN_CROSS_CORRELATION, CUDNN_DATA_FLOAT);
#else
cudnnSetConvolution2dDescriptor(l->convDesc, l->pad, l->pad, l->stride, l->stride, 1, 1, CUDNN_CROSS_CORRELATION);
#endif
#if CUDNN_MAJOR >= 7
cudnnSetConvolutionGroupCount(l->convDesc, l->groups);
#else
if(l->groups > 1){
error("CUDNN < 7 doesn't support groups, please upgrade!");
}
#endif
#if CUDNN_MAJOR >= 8
int returnedAlgoCount;
cudnnConvolutionFwdAlgoPerf_t fw_results[2 * CUDNN_CONVOLUTION_FWD_ALGO_COUNT];
cudnnConvolutionBwdDataAlgoPerf_t bd_results[2 * CUDNN_CONVOLUTION_BWD_DATA_ALGO_COUNT];
cudnnConvolutionBwdFilterAlgoPerf_t bf_results[2 * CUDNN_CONVOLUTION_BWD_FILTER_ALGO_COUNT];
cudnnFindConvolutionForwardAlgorithm(cudnn_handle(),
l->srcTensorDesc,
l->weightDesc,
l->convDesc,
l->dstTensorDesc,
CUDNN_CONVOLUTION_FWD_ALGO_COUNT,
&returnedAlgoCount,
fw_results);
for(int algoIndex = 0; algoIndex < returnedAlgoCount; ++algoIndex){
#if PRINT_CUDNN_ALGO > 0
printf("^^^^ %s for Algo %d: %f time requiring %llu memory\n",
cudnnGetErrorString(fw_results[algoIndex].status),
fw_results[algoIndex].algo, fw_results[algoIndex].time,
(unsigned long long)fw_results[algoIndex].memory);
#endif
if( fw_results[algoIndex].memory < MEMORY_LIMIT ){
l->fw_algo = fw_results[algoIndex].algo;
break;
}
}
cudnnFindConvolutionBackwardDataAlgorithm(cudnn_handle(),
l->weightDesc,
l->ddstTensorDesc,
l->convDesc,
l->dsrcTensorDesc,
CUDNN_CONVOLUTION_BWD_DATA_ALGO_COUNT,
&returnedAlgoCount,
bd_results);
for(int algoIndex = 0; algoIndex < returnedAlgoCount; ++algoIndex){
#if PRINT_CUDNN_ALGO > 0
printf("^^^^ %s for Algo %d: %f time requiring %llu memory\n",
cudnnGetErrorString(bd_results[algoIndex].status),
bd_results[algoIndex].algo, bd_results[algoIndex].time,
(unsigned long long)bd_results[algoIndex].memory);
#endif
if( bd_results[algoIndex].memory < MEMORY_LIMIT ){
l->bd_algo = bd_results[algoIndex].algo;
break;
}
}
cudnnFindConvolutionBackwardFilterAlgorithm(cudnn_handle(),
l->srcTensorDesc,
l->ddstTensorDesc,
l->convDesc,
l->dweightDesc,
CUDNN_CONVOLUTION_BWD_FILTER_ALGO_COUNT,
&returnedAlgoCount,
bf_results);
for(int algoIndex = 0; algoIndex < returnedAlgoCount; ++algoIndex){
#if PRINT_CUDNN_ALGO > 0
printf("^^^^ %s for Algo %d: %f time requiring %llu memory\n",
cudnnGetErrorString(bf_results[algoIndex].status),
bf_results[algoIndex].algo, bf_results[algoIndex].time,
(unsigned long long)bf_results[algoIndex].memory);
#endif
if( bf_results[algoIndex].memory < MEMORY_LIMIT ){
l->bf_algo = bf_results[algoIndex].algo;
break;
}
}
#else
cudnnGetConvolutionForwardAlgorithm(cudnn_handle(),
l->srcTensorDesc,
l->weightDesc,
l->convDesc,
l->dstTensorDesc,
CUDNN_CONVOLUTION_FWD_SPECIFY_WORKSPACE_LIMIT,
2000000000,
&l->fw_algo);
cudnnGetConvolutionBackwardDataAlgorithm(cudnn_handle(),
l->weightDesc,
l->ddstTensorDesc,
l->convDesc,
l->dsrcTensorDesc,
CUDNN_CONVOLUTION_BWD_DATA_SPECIFY_WORKSPACE_LIMIT,
2000000000,
&l->bd_algo);
cudnnGetConvolutionBackwardFilterAlgorithm(cudnn_handle(),
l->srcTensorDesc,
l->ddstTensorDesc,
l->convDesc,
l->dweightDesc,
CUDNN_CONVOLUTION_BWD_FILTER_SPECIFY_WORKSPACE_LIMIT,
2000000000,
&l->bf_algo);
#endif
}
#endif
#endif
convolutional_layer make_convolutional_layer(int batch, int h, int w, int c, int n, int groups, int size, int stride, int padding, ACTIVATION activation, int batch_normalize, int binary, int xnor, int adam)
{
int i;
convolutional_layer l = {0};
l.type = CONVOLUTIONAL;
l.groups = groups;
l.h = h;
l.w = w;
l.c = c;
l.n = n;
l.binary = binary;
l.xnor = xnor;
l.batch = batch;
l.stride = stride;
l.size = size;
l.pad = padding;
l.batch_normalize = batch_normalize;
l.weights = calloc(c/groups*n*size*size, sizeof(float));
l.weight_updates = calloc(c/groups*n*size*size, sizeof(float));
l.biases = calloc(n, sizeof(float));
l.bias_updates = calloc(n, sizeof(float));
l.nweights = c/groups*n*size*size;
l.nbiases = n;
// float scale = 1./sqrt(size*size*c);
float scale = sqrt(2./(size*size*c/l.groups));
//printf("convscale %f\n", scale);
//scale = .02;
//for(i = 0; i < c*n*size*size; ++i) l.weights[i] = scale*rand_uniform(-1, 1);
for(i = 0; i < l.nweights; ++i) l.weights[i] = scale*rand_normal();
int out_w = convolutional_out_width(l);
int out_h = convolutional_out_height(l);
l.out_h = out_h;
l.out_w = out_w;
l.out_c = n;
l.outputs = l.out_h * l.out_w * l.out_c;
l.inputs = l.w * l.h * l.c;
l.output = calloc(l.batch*l.outputs, sizeof(float));
l.delta = calloc(l.batch*l.outputs, sizeof(float));
l.forward = forward_convolutional_layer;
l.backward = backward_convolutional_layer;
l.update = update_convolutional_layer;
if(binary){
l.binary_weights = calloc(l.nweights, sizeof(float));
l.cweights = calloc(l.nweights, sizeof(char));
l.scales = calloc(n, sizeof(float));
}
if(xnor){
l.binary_weights = calloc(l.nweights, sizeof(float));
l.binary_input = calloc(l.inputs*l.batch, sizeof(float));
}
if(batch_normalize){
l.scales = calloc(n, sizeof(float));
l.scale_updates = calloc(n, sizeof(float));
for(i = 0; i < n; ++i){
l.scales[i] = 1;
}
l.mean = calloc(n, sizeof(float));
l.variance = calloc(n, sizeof(float));
l.mean_delta = calloc(n, sizeof(float));
l.variance_delta = calloc(n, sizeof(float));
l.rolling_mean = calloc(n, sizeof(float));
l.rolling_variance = calloc(n, sizeof(float));
l.x = calloc(l.batch*l.outputs, sizeof(float));
l.x_norm = calloc(l.batch*l.outputs, sizeof(float));
}
if(adam){
l.m = calloc(l.nweights, sizeof(float));
l.v = calloc(l.nweights, sizeof(float));
l.bias_m = calloc(n, sizeof(float));
l.scale_m = calloc(n, sizeof(float));
l.bias_v = calloc(n, sizeof(float));
l.scale_v = calloc(n, sizeof(float));
}
#ifdef GPU
l.forward_gpu = forward_convolutional_layer_gpu;
l.backward_gpu = backward_convolutional_layer_gpu;
l.update_gpu = update_convolutional_layer_gpu;
if(gpu_index >= 0){
if (adam) {
l.m_gpu = cuda_make_array(l.m, l.nweights);
l.v_gpu = cuda_make_array(l.v, l.nweights);
l.bias_m_gpu = cuda_make_array(l.bias_m, n);
l.bias_v_gpu = cuda_make_array(l.bias_v, n);
l.scale_m_gpu = cuda_make_array(l.scale_m, n);
l.scale_v_gpu = cuda_make_array(l.scale_v, n);
}
l.weights_gpu = cuda_make_array(l.weights, l.nweights);
l.weight_updates_gpu = cuda_make_array(l.weight_updates, l.nweights);
l.biases_gpu = cuda_make_array(l.biases, n);
l.bias_updates_gpu = cuda_make_array(l.bias_updates, n);
l.delta_gpu = cuda_make_array(l.delta, l.batch*out_h*out_w*n);
l.output_gpu = cuda_make_array(l.output, l.batch*out_h*out_w*n);
if(binary){
l.binary_weights_gpu = cuda_make_array(l.weights, l.nweights);
}
if(xnor){
l.binary_weights_gpu = cuda_make_array(l.weights, l.nweights);
l.binary_input_gpu = cuda_make_array(0, l.inputs*l.batch);
}
if(batch_normalize){
l.mean_gpu = cuda_make_array(l.mean, n);
l.variance_gpu = cuda_make_array(l.variance, n);
l.rolling_mean_gpu = cuda_make_array(l.mean, n);
l.rolling_variance_gpu = cuda_make_array(l.variance, n);
l.mean_delta_gpu = cuda_make_array(l.mean, n);
l.variance_delta_gpu = cuda_make_array(l.variance, n);
l.scales_gpu = cuda_make_array(l.scales, n);
l.scale_updates_gpu = cuda_make_array(l.scale_updates, n);
l.x_gpu = cuda_make_array(l.output, l.batch*out_h*out_w*n);
l.x_norm_gpu = cuda_make_array(l.output, l.batch*out_h*out_w*n);
}
#ifdef CUDNN
cudnnCreateTensorDescriptor(&l.normTensorDesc);
cudnnCreateTensorDescriptor(&l.srcTensorDesc);
cudnnCreateTensorDescriptor(&l.dstTensorDesc);
cudnnCreateFilterDescriptor(&l.weightDesc);
cudnnCreateTensorDescriptor(&l.dsrcTensorDesc);
cudnnCreateTensorDescriptor(&l.ddstTensorDesc);
cudnnCreateFilterDescriptor(&l.dweightDesc);
cudnnCreateConvolutionDescriptor(&l.convDesc);
cudnn_convolutional_setup(&l);
#endif
}
#endif
l.workspace_size = get_workspace_size(l);
l.activation = activation;
fprintf(stderr, "conv %5d %2d x%2d /%2d %4d x%4d x%4d -> %4d x%4d x%4d %5.3f BFLOPs\n", n, size, size, stride, w, h, c, l.out_w, l.out_h, l.out_c, (2.0 * l.n * l.size*l.size*l.c/l.groups * l.out_h*l.out_w)/1000000000.);
return l;
}
void denormalize_convolutional_layer(convolutional_layer l)
{
int i, j;
for(i = 0; i < l.n; ++i){
float scale = l.scales[i]/sqrt(l.rolling_variance[i] + .00001);
for(j = 0; j < l.c/l.groups*l.size*l.size; ++j){
l.weights[i*l.c/l.groups*l.size*l.size + j] *= scale;
}
l.biases[i] -= l.rolling_mean[i] * scale;
l.scales[i] = 1;
l.rolling_mean[i] = 0;
l.rolling_variance[i] = 1;
}
}
/*
void test_convolutional_layer()
{
convolutional_layer l = make_convolutional_layer(1, 5, 5, 3, 2, 5, 2, 1, LEAKY, 1, 0, 0, 0);
l.batch_normalize = 1;
float data[] = {1,1,1,1,1,
1,1,1,1,1,
1,1,1,1,1,
1,1,1,1,1,
1,1,1,1,1,
2,2,2,2,2,
2,2,2,2,2,
2,2,2,2,2,
2,2,2,2,2,
2,2,2,2,2,
3,3,3,3,3,
3,3,3,3,3,
3,3,3,3,3,
3,3,3,3,3,
3,3,3,3,3};
//net.input = data;
//forward_convolutional_layer(l);
}
*/
void resize_convolutional_layer(convolutional_layer *l, int w, int h)
{
l->w = w;
l->h = h;
int out_w = convolutional_out_width(*l);
int out_h = convolutional_out_height(*l);
l->out_w = out_w;
l->out_h = out_h;
l->outputs = l->out_h * l->out_w * l->out_c;
l->inputs = l->w * l->h * l->c;
l->output = realloc(l->output, l->batch*l->outputs*sizeof(float));
l->delta = realloc(l->delta, l->batch*l->outputs*sizeof(float));
if(l->batch_normalize){
l->x = realloc(l->x, l->batch*l->outputs*sizeof(float));
l->x_norm = realloc(l->x_norm, l->batch*l->outputs*sizeof(float));
}
#ifdef GPU
cuda_free(l->delta_gpu);
cuda_free(l->output_gpu);
l->delta_gpu = cuda_make_array(l->delta, l->batch*l->outputs);
l->output_gpu = cuda_make_array(l->output, l->batch*l->outputs);
if(l->batch_normalize){
cuda_free(l->x_gpu);
cuda_free(l->x_norm_gpu);
l->x_gpu = cuda_make_array(l->output, l->batch*l->outputs);
l->x_norm_gpu = cuda_make_array(l->output, l->batch*l->outputs);
}
#ifdef CUDNN
cudnn_convolutional_setup(l);
#endif
#endif
l->workspace_size = get_workspace_size(*l);
}
void add_bias(float *output, float *biases, int batch, int n, int size)
{
int i,j,b;
for(b = 0; b < batch; ++b){
for(i = 0; i < n; ++i){
for(j = 0; j < size; ++j){
output[(b*n + i)*size + j] += biases[i];
}
}
}
}
void scale_bias(float *output, float *scales, int batch, int n, int size)
{
int i,j,b;
for(b = 0; b < batch; ++b){
for(i = 0; i < n; ++i){
for(j = 0; j < size; ++j){
output[(b*n + i)*size + j] *= scales[i];
}
}
}
}
void backward_bias(float *bias_updates, float *delta, int batch, int n, int size)
{
int i,b;
for(b = 0; b < batch; ++b){
for(i = 0; i < n; ++i){
bias_updates[i] += sum_array(delta+size*(i+b*n), size);
}
}
}
void forward_convolutional_layer(convolutional_layer l, network net)
{
int i, j;
fill_cpu(l.outputs*l.batch, 0, l.output, 1);
if(l.xnor){
binarize_weights(l.weights, l.n, l.c/l.groups*l.size*l.size, l.binary_weights);
swap_binary(&l);
binarize_cpu(net.input, l.c*l.h*l.w*l.batch, l.binary_input);
net.input = l.binary_input;
}
int m = l.n/l.groups;
int k = l.size*l.size*l.c/l.groups;
int n = l.out_w*l.out_h;
for(i = 0; i < l.batch; ++i){
for(j = 0; j < l.groups; ++j){
float *a = l.weights + j*l.nweights/l.groups;
float *b = net.workspace;
float *c = l.output + (i*l.groups + j)*n*m;
float *im = net.input + (i*l.groups + j)*l.c/l.groups*l.h*l.w;
if (l.size == 1) {
b = im;
} else {
im2col_cpu(im, l.c/l.groups, l.h, l.w, l.size, l.stride, l.pad, b);
}
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
}
}
if(l.batch_normalize){
forward_batchnorm_layer(l, net);
} else {
add_bias(l.output, l.biases, l.batch, l.n, l.out_h*l.out_w);
}
activate_array(l.output, l.outputs*l.batch, l.activation);
if(l.binary || l.xnor) swap_binary(&l);
}
void backward_convolutional_layer(convolutional_layer l, network net)
{
int i, j;
int m = l.n/l.groups;
int n = l.size*l.size*l.c/l.groups;
int k = l.out_w*l.out_h;
gradient_array(l.output, l.outputs*l.batch, l.activation, l.delta);
if(l.batch_normalize){
backward_batchnorm_layer(l, net);
} else {
backward_bias(l.bias_updates, l.delta, l.batch, l.n, k);
}
for(i = 0; i < l.batch; ++i){
for(j = 0; j < l.groups; ++j){
float *a = l.delta + (i*l.groups + j)*m*k;
float *b = net.workspace;
float *c = l.weight_updates + j*l.nweights/l.groups;
float *im = net.input + (i*l.groups + j)*l.c/l.groups*l.h*l.w;
float *imd = net.delta + (i*l.groups + j)*l.c/l.groups*l.h*l.w;
if(l.size == 1){
b = im;
} else {
im2col_cpu(im, l.c/l.groups, l.h, l.w,
l.size, l.stride, l.pad, b);
}
gemm(0,1,m,n,k,1,a,k,b,k,1,c,n);
if (net.delta) {
a = l.weights + j*l.nweights/l.groups;
b = l.delta + (i*l.groups + j)*m*k;
c = net.workspace;
if (l.size == 1) {
c = imd;
}
gemm(1,0,n,k,m,1,a,n,b,k,0,c,k);
if (l.size != 1) {
col2im_cpu(net.workspace, l.c/l.groups, l.h, l.w, l.size, l.stride, l.pad, imd);
}
}
}
}
}
void update_convolutional_layer(convolutional_layer l, update_args a)
{
float learning_rate = a.learning_rate*l.learning_rate_scale;
float momentum = a.momentum;
float decay = a.decay;
int batch = a.batch;
axpy_cpu(l.n, learning_rate/batch, l.bias_updates, 1, l.biases, 1);
scal_cpu(l.n, momentum, l.bias_updates, 1);
if(l.scales){
axpy_cpu(l.n, learning_rate/batch, l.scale_updates, 1, l.scales, 1);
scal_cpu(l.n, momentum, l.scale_updates, 1);
}
axpy_cpu(l.nweights, -decay*batch, l.weights, 1, l.weight_updates, 1);
axpy_cpu(l.nweights, learning_rate/batch, l.weight_updates, 1, l.weights, 1);
scal_cpu(l.nweights, momentum, l.weight_updates, 1);
}
image get_convolutional_weight(convolutional_layer l, int i)
{
int h = l.size;
int w = l.size;
int c = l.c/l.groups;
return float_to_image(w,h,c,l.weights+i*h*w*c);
}
void rgbgr_weights(convolutional_layer l)
{
int i;
for(i = 0; i < l.n; ++i){
image im = get_convolutional_weight(l, i);
if (im.c == 3) {
rgbgr_image(im);
}
}
}
void rescale_weights(convolutional_layer l, float scale, float trans)
{
int i;
for(i = 0; i < l.n; ++i){
image im = get_convolutional_weight(l, i);
if (im.c == 3) {
scale_image(im, scale);
float sum = sum_array(im.data, im.w*im.h*im.c);
l.biases[i] += sum*trans;
}
}
}
image *get_weights(convolutional_layer l)
{
image *weights = calloc(l.n, sizeof(image));
int i;
for(i = 0; i < l.n; ++i){
weights[i] = copy_image(get_convolutional_weight(l, i));
normalize_image(weights[i]);
/*
char buff[256];
sprintf(buff, "filter%d", i);
save_image(weights[i], buff);
*/
}
//error("hey");
return weights;
}
image *visualize_convolutional_layer(convolutional_layer l, char *window, image *prev_weights)
{
image *single_weights = get_weights(l);
show_images(single_weights, l.n, window);
image delta = get_convolutional_image(l);
image dc = collapse_image_layers(delta, 1);
char buff[256];
sprintf(buff, "%s: Output", window);
//show_image(dc, buff);
//save_image(dc, buff);
free_image(dc);
return single_weights;
}
2.3接下来找个图像测一下
新建一个package，名叫detecting_test，我是放在了darknet_ros同级目录
catkin_create_pkg --rosdistro noetic detecting_image sensor_msgs cv_bridge roscpp std_msgs image_transport
在src文件下创建image_publisher.cpp
#include <ros/ros.h>
#include <opencv2/opencv.hpp>
#include <image_transport/image_transport.h>
#include <cv_bridge/cv_bridge.h>
#include <opencv2/imgcodecs/legacy/constants_c.h>
int main(int argc, char *argv[])
{
ROS_INFO("-------------------hello world---------------------------");
ros::init(argc, argv, "image_publisher");
ros::NodeHandle nh;
image_transport::ImageTransport it(nh);
image_transport::Publisher pub = it.advertise("/camera/image", 1);
cv::Mat image = cv::imread("/home/qiao/darknet_ros_yolov3/loadedtrunk.jpg", CV_LOAD_IMAGE_COLOR);
// cv::Mat image = cv::imread(argv[1], CV_LOAD_IMAGE_COLOR);
sensor_msgs::ImagePtr msg = cv_bridge::CvImage(std_msgs::Header(), "bgr8",image).toImageMsg();
ros::Rate loop_rate(5);
while (nh.ok()) {
pub.publish(msg);
ros::spinOnce();
loop_rate.sleep();
}
return 0;
}
在src文件下创建image_subscriber.cpp
#include <ros/ros.h>
#include <opencv2/opencv.hpp>
#include <image_transport/image_transport.h>
#include <cv_bridge/cv_bridge.h>
void imageCallback(const sensor_msgs::ImageConstPtr& msg)
{
try
{
cv::imshow("view", cv_bridge::toCvShare(msg, "bgr8")->image);
cv::imwrite("/home/qiao/darknet_ros_yolov3/output.png", cv_bridge::toCvShare(msg, "bgr8")->image);
cv::waitKey(10);
}
catch (cv_bridge::Exception& e)
{
ROS_ERROR("Could not convert from '%s' to 'bgr8'.", msg->encoding.c_str());
}
}
int main(int argc, char **argv)
{
ros::init(argc, argv, "image_subscriber");
// std::cout << "argc: " << argc << std::endl;
// std::cout << "argv: " << argv << std::endl;
ros::NodeHandle nh;
cv::namedWindow("view");
//cv::startWindowThread();
image_transport::ImageTransport it(nh);
image_transport::Subscriber sub = it.subscribe("/darknet_ros/detection_image", 1, imageCallback); ///darknet_ros/detection_image
ros::spin();
cv::destroyWindow("view");
return 0;
}
有两点说明:
- 在image_publisher.cpp中，修改了话题的发布名称，和ros.yaml中的camera_reading的topic要一致，才能调用darknet进行检测。检测的图片路径根据自己的路径修改好。
- 注意订阅的话题名称要和ros.yaml中的publisher的detection_image的topic要一致，得到darknet的检测结果并进行存储。
然后运行
roscore
rosrun detecting_image image_publisher
roslaunch darknet_ros darknet_ros.launch
rosrun detecting_image image_subscriber
到这应该是没问题的。
接下来是单步调试的配置:
参考这个链接: VS Code 配置 ROS catkin
3. 插件是按照这个博客配置的。
1.1 C/C++ms-vscode.cpptools
1.2 C++ Intellisenseaustin.code-gnu-global
1.3 CMaketwxs.cmake
1.4 CMake Toolsms-vscode.cmake-tools
1.5 ROS ms-iot.vscode-ros
1.6 catkin-toolsbetwo.b2-catkin-tools
4. 配置文件
c_cpp_properties.json 这里配置C++编译，按照自己项目需求微调"includePath" “compilerPath” “cStandard” “cppStandard” 等参数。
这里的 “configurationProvider” 参数一定要设置为 “b2.catkin_tools”，这样才能够使用 catkin-tools 插件提供的自动补全功能。
{
"configurations": [
{
"name": "ROS",
"includePath": [
"${workspaceFolder}/**",
"/usr/local/include/opencv4",
"/opt/ros/noetic/include/**",
"/usr/include/**",
"/opt/ros/noetic/include"
],
"defines": [],
"compilerPath": "/usr/bin/gcc",
"cStandard": "c11",
"cppStandard": "c++17",
"intelliSenseMode": "linux-gcc-x64",
"configurationProvider": "b2.catkin_tools",
"compileCommands": "${workspaceFolder}/build/compile_commands.json"
// "compileCommands": "${workspaceFolder}/.vscode/build/compile_commands.json"
}
],
"version": 4
}
tasks.json
{
"version": "2.0.0",
"tasks": [
{
"label": "catkin_make",
"type": "shell",
"command": "catkin_make",
"args": [
"-DCMAKE_BUILD_TYPE=Release"
],
"group": {
"kind": "build",
"isDefault": true
},
"presentation": {
"reveal": "always"
},
"problemMatcher": "$msCompile"
},
{
"label": "build",
"type": "shell",
"command": "catkin config --extend /opt/ros/noetic && catkin build -DCMAKE_BUILD_TYPE = Debug -j16",
"group": {
"kind": "build",
"isDefault": true
},
"problemMatcher": []
},
{
"label": "clean",
"type": "shell",
"command": "catkin clean --yes"
},
{
"label": "release",
"type": "shell",
"command": "sudo checkinstall --install=no catkin build -j16 --cmake--args -DCMAKE_BUILD_TYPE=Release"
},
{
"label": "make_debug",
"type": "catkin_make",
"args": [
"--directory",
"${workspaceFolder}",
"-DCMAKE_BUILD_TYPE=Debug"
],
"problemMatcher": [
"$catkin-gcc"
],
"group": {
"kind": "build",
"isDefault": true
}
},
{
"label": "make_debug_and_core",
"dependsOn": [
"make_debug",
"roscore: roscore"
],
"problemMatcher": [
"$catkin-gcc"
]
},
{
"label": "prerun",
"type": "shell",
"command": "source ${workspaceFolder}/devel/setup.bash"
},
]
}
launch.json
{
// Use IntelliSense to learn about possible attributes.
// Hover to view descriptions of existing attributes.
// For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
"version": "0.2.0",
"configurations": [
// {
// "name": "(gdb) Launch",
// "type": "cppdbg",
// "request": "launch",
// "program": "${workspaceFolder}/devel/lib/detecting_image/image_publisher", // start cpp file
// "args": [],
// "stopAtEntry": false,
// "cwd": "${workspaceFolder}",
// "environment": [],
// "externalConsole": false,
// "MIMode": "gdb",
// "preLaunchTask": "catkin_make",
// "setupCommands": [
// {
// "description": "Enable pretty-printing for gdb",
// "text": "-enable-pretty-printing",
// "ignoreFailures": true
// }
// ]
// },
{
"name": "ROS: Launch",
"type": "ros",
"request": "launch",
"target": "${workspaceFolder}/src/darknet_ros/darknet_ros/launch/darknet_ros.launch",
"preLaunchTask": "make_debug_and_core",
}
]
}
另外一个setting.json不需要处理，我没管。
第一次编译需要:
catkin_make -DCMAKE_BUILD_TYPE=Release -DCMAKE_EXPORT_COMPILE_COMMANDS=Yes
配置完重启vscode就可以了。
之后就不需要了。
注意：
1 我调试image_publisher的时候直接调试那个cpp文件
2 调试darknet_launch.launch里的cpp是先终端启动发送话题并带一张图像，再将launch.json里启动publisher的字典注释掉，再启动的launch文件的，这样是可以debug的
其他的参考链接:
https://blog.csdn.net/JIEJINQUANIL/article/details/107986662
https://zhuanlan.zhihu.com/p/150331361?from_voters_page=true
https://blog.csdn.net/ABC_ORANGE/article/details/102665792
扫一扫
4600
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
