YOLO-V3訓練中會遇到的問題

原創 凌空的桨 2018-11-23 06:30

0.數據集的準備，採用voc格式的數據集，數據集格式

VOCdevkit
 -VOC2007
   -JPEGImages
   -Annotations
   -ImageSets
     -Layout
     -Main
     -Segmentation

就是VOCdevkit裏面包含VOC2007；VOC2007裏面包含Annotations，ImageSets，JPEGImages文件夾；ImageSets裏面有Layout，Main，Segmentation文件夾。

其中Annotations是你的標籤文件（就是.xml文件）；JPEGImages文件夾裏面是你的圖片；ImageSets裏面可以什麼都不放。用不上。

1.將.xml轉換成.txt文件

linux在script裏面，windows在voc文件夾裏面，有個叫voc_label.py文件，它可以進行轉換。

可能會遇到問題，比如你的圖片是00001.jpg開始的，它就會報錯，說00000.jpg不存在。這時候需要將voc_label.py代碼的第53行添一句，將00000去掉

image_ids=image_ids[1:]

整體的代碼就像這樣，我是分一類物體，物體類別就是1，只含有訓練集。

import xml.etree.ElementTree as ET
import pickle
import os
from os import listdir, getcwd
from os.path import join

# sets=[('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test')]

# classes = ["aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"]

sets=[('2007', 'train')]
classes = ["ID"]

def convert(size, box):
    dw = 1./size[0]
    dh = 1./size[1]
    x = (box[0] + box[1])/2.0
    y = (box[2] + box[3])/2.0
    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))
    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()
    image_ids=image_ids[1:]
    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()

如果你是00000.jpg開頭的，可能會遇到得到的.txt標籤爲空。這個時候有兩種解決方案，要麼可以將所有的圖像和對應的標籤文件.xml都以00001開頭。

要麼就將最後一個去掉image_ids=image_ids[:612]，假設你的圖像只有612張。不過這種改法還沒有驗證第一種方法已經得到驗證。

2.然後進行訓練

比如你要訓練yolov3-voc.cfg

你就要修改它：

[net]
# Testing
 #batch=64
 #subdivisions=64
# 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
policy=steps
steps=40000,45000
scales=.1,.1



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

# Downsample

[convolutional]
batch_normalize=1
filters=64
size=3
stride=2
pad=1
activation=leaky

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

[convolutional]
batch_normalize=1
filters=64
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

# Downsample

[convolutional]
batch_normalize=1
filters=128
size=3
stride=2
pad=1
activation=leaky

[convolutional]
batch_normalize=1
filters=64
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

[convolutional]
batch_normalize=1
filters=64
size=1
stride=1
pad=1
activation=leaky

[convolutional]
batch_normalize=1
filters=128
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

# Downsample

[convolutional]
batch_normalize=1
filters=256
size=3
stride=2
pad=1
activation=leaky

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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear


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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear

# Downsample

[convolutional]
batch_normalize=1
filters=512
size=3
stride=2
pad=1
activation=leaky

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

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

[shortcut]
from=-3
activation=linear


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

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

[shortcut]
from=-3
activation=linear


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

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

[shortcut]
from=-3
activation=linear


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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear


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

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

[shortcut]
from=-3
activation=linear


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

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

[shortcut]
from=-3
activation=linear

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

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

[shortcut]
from=-3
activation=linear

# Downsample

[convolutional]
batch_normalize=1
filters=1024
size=3
stride=2
pad=1
activation=leaky

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

[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

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

[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

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

[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

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

[convolutional]
batch_normalize=1
filters=1024
size=3
stride=1
pad=1
activation=leaky

[shortcut]
from=-3
activation=linear

######################

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

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=1024
activation=leaky

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

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=1024
activation=leaky

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

[convolutional]
batch_normalize=1
size=3
stride=1
pad=1
filters=1024
activation=leaky

[convolutional]
size=1
stride=1
pad=1
filters=18  #3*(classes+4+1)
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=1
num=9
jitter=.3
ignore_thresh = .5
truth_thresh = 1
random=0

[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=18 # 3*(classes+4+1)
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=1
num=9
jitter=.3
ignore_thresh = .5
truth_thresh = 1
random=0

[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=18 #3*(classes+4+1)
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=1
num=9
jitter=.3
ignore_thresh = .5
truth_thresh = 1
random=0

這裏需要注意的是 classes和classes上面的filters，filters=3*（classes+4+1）我的是1；

還需要改的是data文件夾裏的voc.data和voc.names；

voc.data:

#voc.data
classes= 1
train  = data/voc/2007_train.txt
#valid  = data/2007_test.txt
#difficult = data/difficult_2007_test.txt
names = data/voc.names
backup = backup/

 voc.names:

ID

在uabntu下非常容易訓練，這裏不再贅述，在windows下，有預訓練文件可以輸入

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

沒有預訓練文件，輸入

darknet_no_gpu.exe detector train data/voc.data cfg/yolov3-voc.cfg

3.最後就是等着結果了

 4.訓練的權重保存在backups裏，前1000次每100次保存一次權重，前10000次每1000次保存一次權重

(0,1000)
    -100
	-200
	.
	.
	.
	-900
(1000,10000)
    -1000
	-2000
	.
	.
	.
	-9000
  .
  .
  .

像這樣，依此類推。

發表評論
所有評論
還沒有人評論，想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章

三、訓練自己的yolo3模型

yolo3整體的文件夾構架如下： 本文使用VOC格式進行訓練。 訓練前將 標籤文件(.xml) 放在 VOCdevkit 文件夾下的 VOC2007 文件夾下的 Annotation 中。 訓練前將 圖片文件 放在 VOCdevkit

无止境x 2020-07-07 15:53:01

Yolov4模型訓練規則和技巧

Yolov4模型訓練技巧總結 Training Yolo v4 (and v3): AlexeyAB官方github介紹: 訓練對應的yolov4.cfg，cfg/yolov4-custom.cfg，cfg/yolov4-tiny

胡椒面er 2020-07-08 12:10:55

YOLOv3訓練實踐（把代碼整合一下，寫個直接xml到txt的代碼，xml到csv,xml到record的代碼 留坑）

1.準備數據集 1.找足夠多的圖片。 2.下載labelImg軟件，網址https://tzutalin.github.io/labelImg/，需要翻牆。下載windows最新版。 win10，第一次運行labelImg沒毛病，之後再打

_g_y_ 2020-07-08 10:06:47

YOLOv3 小功能

目錄 1. 僅輸出指定類別的檢測結果 2. 裁剪出目標 3. 添加目標置信度 4. 批量測試圖片並保存至指定文件夾 5. 保存視頻檢測結果 6. 修改模型保存頻率 7. 保存訓練日誌，參數可視化 1. 僅輸出指定類別的檢測結果 coco模

_g_y_ 2020-07-08 10:06:46

二、YOLOv3-訓練部分

二、訓練部分 1、計算loss所需參數 在計算loss的時候，實際上是y_pre和y_true之間的對比：y_pre 就是一幅圖像經過網絡之後的輸出，內部含有三個特徵層的內容；其需要解碼才能夠在圖上作畫y_true 就是一個真實圖像中，

无止境x 2020-07-07 15:53:00

Mask R-CNN tensorflow 訓練自己的數據

Mask R-CNN tensorflow 訓練自己的數據 https://blog.csdn.net/doudou_here/article/details/87855273   labelme https://blog.csdn.ne

小新识图 2020-07-08 12:33:38

深度學習_目標檢測_FPN論文詳解

FPN的創新點 多層特徵 特徵融合 解決了目標檢測中的多尺度問題，通過簡單的網絡連接改變，在基本不增加原有模型計算量的情況下，大幅度提升小物體（small object）的檢測性能。 在物體檢測裏面，有限計算量情況下，網絡的深

CV-GANRocky 2020-07-08 11:57:07

深度學習_目標檢測_“YOLOv5”詳解（持續更新）

YOLOv5可以方便的進行工程化部署： YOLOv5（PyTorch）−>ONNX−>CoreML−>iosYOLOv5（PyTorch） -> ONNX -> CoreML -> iosYOLOv5（PyTorch）−>ONNX

CV-GANRocky 2020-07-08 11:57:07

IOU計算代碼

import numpy as np def compute_iou(box1, box2, standard_coordinates=True): ''' :param box1: [Xmin, Ymin, Xmax

沐雨金鳞 2020-07-08 10:30:45

Faster RCNN圖解

高清圖：faster rcnn流程圖.png 

沐雨金鳞 2020-07-08 10:30:45

NMS計算代碼

''' # INPUT：所有預測出的bounding box (bbx)信息（座標和置信度confidence），　IOU閾值（大於該閾值的bbx將被移除） for object in all objects: (1) 獲取當前目

沐雨金鳞 2020-07-08 10:30:45

yolov4訓練自己的數據集實現安全帽佩戴檢測

昨兒立下來的flag，今天還是要含淚完成的，抓緊時間趕呀！！！ 本次用yolov4來實現人佩戴安全帽檢測，若未佩戴安全帽則將人臉框出來，若佩戴安全帽，則將安全帽以及人臉框出來，多說無益，直接看效果吧！！！ 效果還是不錯的，那

mind_programmonkey 2020-07-08 09:29:00

YOLOV5目標檢測----上手

YOLOV5目標檢測----上手 2020年4月23日，YOLOV4開源 2020年6月10日，YOLOV5開源 這兩個日子，可是把CV類公衆號的運營者忙壞了，各大公衆號都在忙着搶“首發” 當然，YOLOV4是貨真價實的YOLO

西红柿爱吃小番茄 2020-07-08 07:02:00

目標檢測框架——Dtectron2

前言 目標檢測的模型還是很多的，要挨個挨個學還是有點喫力（精力旺盛可忽略），所以這裏總結一下當前流行的目標檢測的框架：darknet yolov4，mmdetection2.0，detectron2.0。至於這三個框架怎麼去選擇，

不会写作文的李华 2020-07-08 06:19:37

目標檢測框架——mmdetection2.0

前言 目標檢測的模型還是很多的，要挨個挨個學還是有點喫力（精力旺盛可忽略），所以這裏總結一下當前流行的目標檢測的框架：darknet yolov4，mmdetection2.0，detectron2.0。至於這三個框架怎麼去選擇，

不会写作文的李华 2020-07-08 06:19:36