darknet-訓練本身的yolov3模型

時間  2019-12-05
標籤 darknet 訓練 本身 yolov3 yolov 模型 简体版
原文   原文鏈接

目錄html

Yolo v3的使用方法

參考自@zhaonanpython

安裝darknet

  • 下載庫文件
git clone https://github.com/pjreddie/darknet
cd darknet
  • 修改Makefile
GPU=1  #0或1
CUDNN=1  #0或1
OPENCV=0  #0或1
OPENMP=0
DEBUG=0
  • 編譯
make
  • 下載預訓練模型
wget https://pjreddie.com/media/files/yolov3.weights
  • 用預訓練模型進行簡單的測試
./darknet detect cfg/yolov3.cfg yolov3.weights data/dog.jpg

訓練Pascal VOC格式的數據

  • 生成Labels，由於darknet不須要xml文件，須要.txt文件（格式： ）

用voc_label.py（位於./scripts）cat voc_label.py 共修改四處git

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join
sets=[('2007', 'train'), ('2007', 'val'), ('2007', 'test')]  #替換爲本身的數據集
classes = ["head", "eye", "nose"]     #修改成本身的類別

def convert(size, box):
    dw = 1./(size[0])
    dh = 1./(size[1])
    x = (box[0] + box[1])/2.0 - 1
    y = (box[2] + box[3])/2.0 - 1
    w = box[1] - box[0]
    h = box[3] - box[2]
    x = x*dw
    w = w*dw
    y = y*dh
    h = h*dh
    return (x,y,w,h)
def convert_annotation(year, image_id):
    in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id))  #將數據集放於當前目錄下
    out_file = open('VOCdevkit/VOC%s/labels/%s.txt'%(year, image_id), 'w')
    tree=ET.parse(in_file)
    root = tree.getroot()
    size = root.find('size')
    w = int(size.find('width').text)
    h = int(size.find('height').text)
    for obj in root.iter('object'):
        difficult = obj.find('difficult').text
        cls = obj.find('name').text
        if cls not in classes or int(difficult)==1:
            continue
        cls_id = classes.index(cls)
        xmlbox = obj.find('bndbox')
        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text), float(xmlbox.find('ymax').text))
        bb = convert((w,h), b)
        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')
wd = getcwd()
for year, image_set in sets:
    if not os.path.exists('VOCdevkit/VOC%s/labels/'%(year)):
        os.makedirs('VOCdevkit/VOC%s/labels/'%(year))
    image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split()
    list_file = open('%s_%s.txt'%(year, image_set), 'w')
    for image_id in image_ids:
        list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%(wd, year, image_id))
        convert_annotation(year, image_id)
    list_file.close()   
os.system("cat 2007_train.txt 2007_val.txt > train.txt")     #修改成本身的數據集用做訓練
wget https://pjreddie.com/media/files/voc_label.py
python voc_label.py

在VOCdevkit/VOC2007/labels/中：github

learner@learner-pc:~/darknet/scripts$ ls
2007_test.txt #0   dice_label.sh        imagenet_label.sh  VOCdevkit_original
2007_train.txt #1  gen_tactic.sh        train.txt #3        voc_label.py
2007_val.txt #2 get_coco_dataset.sh  VOCdevkit

這時darknet須要一個txt文件，其中包含了全部的圖片網絡

cat 2007_train.txt 2007_val.txt 2012_*.txt > train.txt

修改cfg文件中的voc.data

classes= 3    #修改成本身的類別數
train  = /home/learner/darknet/data/voc/train.txt   #修改成本身的路徑 or /home/learner/darknet/scripts/2007_test.txt
valid  = /home/learner/darknet/data/voc/2007_test.txt   #修改成本身的路徑 or /home/learner/darknet/scripts/2007_test.txt
names = /home/learner/darknet/data/voc.names  #修改見voc.names
backup = /home/learner/darknet/backup   #修改成本身的路徑，輸出的權重信息將存儲其內

修改VOC.names

head  #本身須要探測的類別，一行一個
eye
nose

下載預訓練卷積層權重

wget https://pjreddie.com/media/files/darknet53.conv.74

修改cfg/yolov3-voc.cfg

[net]
# Testing
 batch=64
 subdivisions=32   #每批訓練的個數=batch/subvisions，根據本身GPU顯存進行修改，顯存不夠改大一些
# Training
# batch=64
# subdivisions=16
width=416
height=416
channels=3
momentum=0.9
decay=0.0005
angle=0
saturation = 1.5
exposure = 1.5
hue=.1

learning_rate=0.001
burn_in=1000
max_batches = 50200  #訓練步數
policy=steps
steps=40000,45000  #開始衰減的步數
scales=.1,.1



[convolutional]
batch_normalize=1
filters=32
size=3
stride=1
pad=1
activation=leaky

.....

[convolutional]
size=1
stride=1
pad=1
filters=24   #filters = 3 * ( classes + 5 )   here,filters=3*(3+5)
activation=linear

[yolo]
mask = 6,7,8
anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
classes=3    #修改成本身的類別數
num=9
jitter=.3
ignore_thresh = .5
truth_thresh = 1
random=1

[route]
layers = -4

[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky

[upsample]
stride=2

[route]
layers = -1, 61



[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky

[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky

[convolutional]
batch_normalize=1
filters=256
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=512
activation=leaky

[convolutional]
size=1
stride=1
pad=1
filters=24    #filters = 3 * ( classes + 5 )   here,filters=3*(3+5)
activation=linear

[yolo]
mask = 3,4,5
anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
classes=3  #修改成本身的類別數
num=9
jitter=.3
ignore_thresh = .5
truth_thresh = 1
random=1

[route]
layers = -4

[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky

[upsample]
stride=2

[route]
layers = -1, 36



[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky

[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky

[convolutional]
batch_normalize=1
filters=128
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=256
activation=leaky

[convolutional]
size=1
stride=1
pad=1
filters=24    #filters = 3 * ( classes + 5 )   here,filters=3*(3+5)
activation=linear

[yolo]
mask = 0,1,2
anchors = 10,13,  16,30,  33,23,  30,61,  62,45,  59,119,  116,90,  156,198,  373,326
classes=3   #修改成本身的類別數
num=9
jitter=.3
ignore_thresh = .5
truth_thresh = 1
random=1

訓練本身的模型

1 單GPU訓練：./darknet -i <gpu_id> detector train <data_cfg> <train_cfg> <weights>dom

./darknet detector train cfg/voc.data cfg/yolov3-voc.cfg darknet53.conv.74

2 多GPU訓練，格式爲0,1,2,3：./darknet detector train <data_cfg> <model_cfg> <weights> -gpus <gpu_list>ide

./darknet detector train cfg/voc.data cfg/yolov3-voc.cfg darknet53.conv.74 -gpus 0,1,2,3

測試Yolo模型

測試單張圖片：

  • 測試單張圖片，須要編譯時有OpenCV支持：./darknet detector test <data_cfg> <test_cfg> <weights> <image_file> #本次測試無opencv支持
  • <test_cfg>文件中batch和subdivisions兩項必須爲1。
  • 測試時還能夠用-thresh和-hier選項指定對應參數。

  • ./darknet detector test cfg/voc.data cfg/yolov3-voc.cfg backup/yolov3-voc_20000.weights Eminem.jpg函數

    批量測試圖片

    • yolov3-voc.cfg(cfg文件夾下)文件中batch和subdivisions兩項必須爲1。測試

    • 在detector.c中增長頭文件：優化

      #include <unistd.h>  /* Many POSIX functions (but not all, by a large margin) */
#include <fcntl.h>   /* open(), creat() - and fcntl() */

  • 在前面添加GetFilename(char p)函數

    #include "darknet.h"
#include <sys/stat.h>  //需增長的頭文件
#include<stdio.h>
#include<time.h>
#include<sys/types.h>  //需增長的頭文件
static int coco_ids[] = {1,2,3,4,5,6,7,8,9,10,11,13,14,15,16,17,18,19,20,21,22,23,24,25,27,28,31,32,33,34,35,36,37,38,39,40,41,42,43,44,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,67,70,72,73,74,75,76,77,78,79,80,81,82,84,85,86,87,88,89,90};

char *GetFilename(char *p)
{ 
    static char name[30]={""};
    char *q = strrchr(p,'/') + 1;
    strncpy(name,q,20);
    return name;
}

  • 用下面代碼替換detector.c文件（example文件夾下）的void test_detector函數（注意有3處要改爲本身的路徑）

void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh, float hier_thresh, char *outfile, int fullscreen)
{
    list *options = read_data_cfg(datacfg);
    char *name_list = option_find_str(options, "names", "data/names.list");
    char **names = get_labels(name_list);
 
    image **alphabet = load_alphabet();
    network *net = load_network(cfgfile, weightfile, 0);
    set_batch_network(net, 1);
    srand(2222222);
    double time;
    char buff[256];
    char *input = buff;
    float nms=.45;
    int i=0;
    while(1){
        if(filename){
            strncpy(input, filename, 256);
            image im = load_image_color(input,0,0);
            image sized = letterbox_image(im, net->w, net->h);
        //image sized = resize_image(im, net->w, net->h);
        //image sized2 = resize_max(im, net->w);
        //image sized = crop_image(sized2, -((net->w - sized2.w)/2), -((net->h - sized2.h)/2), net->w, net->h);
        //resize_network(net, sized.w, sized.h);
            layer l = net->layers[net->n-1];
 
 
            float *X = sized.data;
            time=what_time_is_it_now();
            network_predict(net, X);
            printf("%s: Predicted in %f seconds.\n", input, what_time_is_it_now()-time);
            int nboxes = 0;
            detection *dets = get_network_boxes(net, im.w, im.h, thresh, hier_thresh, 0, 1, &nboxes);
            //printf("%d\n", nboxes);
            //if (nms) do_nms_obj(boxes, probs, l.w*l.h*l.n, l.classes, nms);
            if (nms) do_nms_sort(dets, nboxes, l.classes, nms);
                draw_detections(im, dets, nboxes, thresh, names, alphabet, l.classes);
                free_detections(dets, nboxes);
            if(outfile)
             {
                save_image(im, outfile);
             }
            else{
                save_image(im, "predictions");
#ifdef OPENCV
                cvNamedWindow("predictions", CV_WINDOW_NORMAL); 
                if(fullscreen){
                cvSetWindowProperty("predictions", CV_WND_PROP_FULLSCREEN, CV_WINDOW_FULLSCREEN);
                }
                show_image(im, "predictions",0);
                cvWaitKey(0);
                cvDestroyAllWindows();
#endif
            }
            free_image(im);
            free_image(sized);
            if (filename) break;
         } 
        else {
            printf("Enter Image Path: ");
            fflush(stdout);
            input = fgets(input, 256, stdin);
            if(!input) return;
            strtok(input, "\n");
   
            list *plist = get_paths(input);
            char **paths = (char **)list_to_array(plist);
             printf("Start Testing!\n");
            int m = plist->size;
            if(access("/home/learner/darknet/data/outv3tiny_dpj",0)==-1)//"/home/learner/darknet/data"修改爲本身的路徑
            {
              if (mkdir("/home/learner/darknet/data/outv3tiny_dpj",0777))//"/home/learner/darknet/data"修改爲本身的路徑
               {
                 printf("creat file bag failed!!!");
               }
            }
            for(i = 0; i < m; ++i){
             char *path = paths[i];
             image im = load_image_color(path,0,0);
             image sized = letterbox_image(im, net->w, net->h);
        //image sized = resize_image(im, net->w, net->h);
        //image sized2 = resize_max(im, net->w);
        //image sized = crop_image(sized2, -((net->w - sized2.w)/2), -((net->h - sized2.h)/2), net->w, net->h);
        //resize_network(net, sized.w, sized.h);
        layer l = net->layers[net->n-1];
 
 
        float *X = sized.data;
        time=what_time_is_it_now();
        network_predict(net, X);
        printf("Try Very Hard:");
        printf("%s: Predicted in %f seconds.\n", path, what_time_is_it_now()-time);
        int nboxes = 0;
        detection *dets = get_network_boxes(net, im.w, im.h, thresh, hier_thresh, 0, 1, &nboxes);
        //printf("%d\n", nboxes);
        //if (nms) do_nms_obj(boxes, probs, l.w*l.h*l.n, l.classes, nms);
        if (nms) do_nms_sort(dets, nboxes, l.classes, nms);
        draw_detections(im, dets, nboxes, thresh, names, alphabet, l.classes);
        free_detections(dets, nboxes);
        if(outfile){
            save_image(im, outfile);
        }
        else{
             
             char b[2048];
            sprintf(b,"/home/learner/darknet/data/outv3tiny_dpj/%s",GetFilename(path));//"/home/leaner/darknet/data"修改爲本身的路徑
            
            save_image(im, b);
            printf("save %s successfully!\n",GetFilename(path));
/*
#ifdef OPENCV
            //cvNamedWindow("predictions", CV_WINDOW_NORMAL); 
            if(fullscreen){
                cvSetWindowProperty("predictions", CV_WND_PROP_FULLSCREEN, CV_WINDOW_FULLSCREEN);
            }
            //show_image(im, "predictions");
            //cvWaitKey(0);
            //cvDestroyAllWindows();
#endif*/
        }
 
        free_image(im);
        free_image(sized);
        if (filename) break;
        }
      }
    }
}
  • 從新進行編譯
make clean
make
  • 開始批量測試
./darknet detector test cfg/voc.data cfg/yolov3-voc.cfg backup/yolov3-voc_20000.weights
  • 輸入Image Path(全部的測試文件的路徑，能夠複製voc.data中valid後邊的路徑):
/home/learner/darknet/data/voc/2007_test.txt # 完整路徑
  • 結果都保存在./data/out文件夾下

生成預測結果

生成預測結果

  • ./darknet detector valid <data_cfg> <test_cfg> <weights> <out_file>
  • yolov3-voc.cfg(cfg文件夾下)文件中batch和subdivisions兩項必須爲1。
  • 結果生成在<data_cfg>的results指定的目錄下以<out_file>開頭的若干文件中，若<data_cfg>沒有指定results，那麼默認爲<darknet_root>/results。
  • 執行語句以下：在終端只返回用時，在./results/comp4_det_test_[類名].txt裏保存測試結果
./darknet detector valid cfg/voc.data cfg/yolov3-voc.cfg backup/yolov3-voc_20000.weights

採用第三方compute_mAP

下載第三方庫：

git clone https://github.com/LianjiLi/yolo-compute-map.git

進行以下修改:

  • 修改darknet/examples/detector.c中validate_detector()

    char *valid_images = option_find_str(options, "valid", "./data/2007_test.txt");//改爲本身的測試文件路徑

if(!outfile) outfile = "comp4_det_test_";
        fps = calloc(classes, sizeof(FILE *));
        for(j = 0; j < classes; ++j){
            snprintf(buff, 1024, "%s/%s.txt", prefix, names[j]);//刪除outfile參數以及對應的%s
            fps[j] = fopen(buff, "w");

  • 從新編譯

    make clean
make

  • 運行valid

    darknet文件夾下運行./darknet detector valid cfg/voc.data cfg/yolov3-tiny.cfg backup/yolov3-tiny_164000.weights（改成本身的模型路徑）

  • 在本文件夾下運行python compute_mAP.py

  • 說明：compute_mAP.py中的test.txt文件內容只有文件名字，不帶絕對路徑，不帶後綴

高級進階

darknet的淺層特徵可視化請參看：http://www.javashuo.com/article/p-gwtgqlbj-ka.html

AlexyAB大神總結的優化經驗請參看：http://www.javashuo.com/article/p-vdejxxdf-ew.html

如何使用Darknet進行分類請參看：http://www.javashuo.com/article/p-oybnqtcr-bv.html

Darknet loss可視化軟件請參看：http://www.javashuo.com/article/p-ovhillwt-gm.html

如何設計更改YOLO網絡結構：https://pprp.github.io/2018/09/20/tricks.html

YOLO詳細改進總結：https://pprp.github.io/2018/06/20/yolo.html

ps: 以上都是本身科研過程當中總結內容，可能不夠系統，歡迎留言討論

Reference

YOLOv3目標檢測總結

官方網站

思路整理自@zhaonan

轉載請註明做者 ^_^, 若有問題請留言。

相關文章
相關標籤/搜索
每日一句
    每一个你不满意的现在，都有一个你没有努力的曾经。
本站公眾號
   歡迎關注本站公眾號,獲取更多信息
聯繫我們 最近搜索 最新文章 沪ICP备13005482号-10 MyBatis教程 SQL 教程 MySQL教程 Java 教程 Thymeleaf 教程 Hibernate教程 Spring教程 Redis教程