Using own data and add write csv in dir predict

README.md

@@ -1,3 +1,18 @@
+## 蛇足
+
+我在bubbliiiing的代码中加入了
+
+import os
+
+os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
+
+解决了我的Initializing libiomp5md.dll, but found libiomp5md.dll already initialized.的报错问题
+
+同时加入了coco转voc数据集的代码。
+
+
+
+
 ## YOLOV5：You Only Look Once目标检测模型在pytorch当中的实现（edition v5.0 in Ultralytics）
 ---
 

coco2voc.py

@@ -0,0 +1,122 @@
+import json
+from xml.dom.minidom import *
+import xml.etree.ElementTree as ET
+import numpy as np
+
+#-----------------------------------------------------------#
+#
+#             本文件用于将coco数据集中的json文件
+#                转化为voc数据集用的xml标签
+#-----------------------------------------------------------#
+# 美化xml文件
+def pretty_xml(element, indent, newline, level=0):  # elemnt为传进来的Elment类，参数indent用于缩进，newline用于换行
+    if element:  # 判断element是否有子元素
+        if (element.text is None) or element.text.isspace():  # 如果element的text没有内容
+            element.text = newline + indent * (level + 1)
+        else:
+            element.text = newline + indent * (level + 1) + element.text.strip() + newline + indent * (level + 1)
+            # else:  # 此处两行如果把注释去掉，Element的text也会另起一行
+            # element.text = newline + indent * (level + 1) + element.text.strip() + newline + indent * level
+    temp = list(element)  # 将element转成list
+    for subelement in temp:
+        if temp.index(subelement) < (len(temp) - 1):  # 如果不是list的最后一个元素，说明下一个行是同级别元素的起始，缩进应一致
+            subelement.tail = newline + indent * (level + 1)
+        else:  # 如果是list的最后一个元素， 说明下一行是母元素的结束，缩进应该少一个
+            subelement.tail = newline + indent * level
+        pretty_xml(subelement, indent, newline, level=level + 1)  # 对子元素进行递归操作
+
+# 写入xml文件
+def write_xml(img_name, width, height, object_dicts, save_path, folder='XJTU_dataset'):
+    '''
+    object_dict = {'name': classes[int(object_category)],
+                            'truncated': int(truncation),
+                            'difficult': int(occlusion),
+                            'xmin':int(bbox_left),
+                            'ymin':int(bbox_top),
+                            'xmax':int(bbox_left) + int(bbox_width),
+                            'ymax':int(bbox_top) + int(bbox_height)
+                            }
+    '''
+    doc = Document
+    root = ET.Element('Annotation')
+    ET.SubElement(root, 'folder').text = folder
+    ET.SubElement(root, 'filename').text = img_name
+    size_node = ET.SubElement(root, 'size')
+    ET.SubElement(size_node, 'width').text = str(width)
+    ET.SubElement(size_node, 'height').text = str(height)
+    ET.SubElement(size_node, 'depth').text = '3'
+    for object_dict in object_dicts:
+        object_node = ET.SubElement(root, 'object')
+        ET.SubElement(object_node, 'name').text = object_dict['name']
+        ET.SubElement(object_node, 'pose').text = 'Unspecified'
+        ET.SubElement(object_node, 'truncated').text = str(object_dict['truncated'])
+        ET.SubElement(object_node, 'difficult').text = str(object_dict['difficult'])
+        bndbox_node = ET.SubElement(object_node, 'bndbox')
+        ET.SubElement(bndbox_node, 'xmin').text = str(object_dict['xmin'])
+        ET.SubElement(bndbox_node, 'ymin').text = str(object_dict['ymin'])
+        ET.SubElement(bndbox_node, 'xmax').text = str(object_dict['xmax'])
+        ET.SubElement(bndbox_node, 'ymax').text = str(object_dict['ymax'])
+
+    pretty_xml(root, '\t', '\n')
+    tree = ET.ElementTree(root)
+    tree.write(save_path, encoding='utf-8')
+
+if __name__ == '__main__':
+    # class_names = ['road', 'car', 'motorcycle', 'person']
+    # f = open("../../../cow/train.json", encoding='utf-8')
+    f = open("cowboyoutfits/train.json", encoding='utf-8')
+    data = json.load(f)
+    annotations = data['annotations']
+    images = data['images']
+    cate = {
+        "87": "belt",
+        "1034": "sunglasses",
+        "131": "boot",
+        "318": "cowboy_hat",
+        "588": "jacket"
+    }
+    images_num = len(images)
+    # print()
+    bbox_infos = {}
+    for image in images:
+        print()
+        bbox_infos[str(image["id"])] = []
+
+    for ann in annotations:
+        print(ann)
+        image_id = ann['image_id']
+        bbox = ann['bbox']
+        bbox_x = bbox[0]
+        bbox_y = bbox[1]
+        bbox_w = bbox[2]
+        bbox_h = bbox[3]
+        class_id = ann['category_id']
+        object_dict = {'name': cate[str(class_id)],
+                       'truncated': 0,
+                       'difficult': 0,
+                       'xmin': int(bbox_x),
+                       'ymin': int(bbox_y),
+                       'xmax': int(bbox_x+bbox_w),
+                       'ymax': int(bbox_y+bbox_h)
+                       }
+        print(object_dict)
+        bbox_infos[str(image_id)].append(object_dict)
+
+
+    txt_results = []
+    for image in images:
+        print(image)
+        image_height = image['height']
+        image_width = image['width']
+        image_id = image['id']
+        image_file_name = image['file_name']
+        object_dicts = bbox_infos[str(image_id)]
+        xml_file_name = image_file_name.strip('.jpg') + '.xml'
+        txt_results.append(image_file_name.strip('.jpg'))
+        # print(image_file_name, image_width, image_height, object_dicts, )
+        write_xml(image_file_name, image_width, image_height, object_dicts, "D:/yolov5-pytorch/cowboy_voc_xml/"+xml_file_name)
+
+    # np.savetxt()
+    # np.savetxt('train.txt', txt_results, fmt="%s", delimiter="\n")
+    # for image in images:
+    #     print(image)

data_analysis.ipynb

@@ -0,0 +1,150 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import tensorboard"
+   ]
+  },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## 批量用yolo将bounding box保存csv\n",
+    "\n",
+    "1.保存bounding box到csv\n",
+    "\n",
+    "2.将代码缩减成一到两个py文件\n",
+    "\n",
+    "3.将bounding box保存到json文件\n",
+    "\n",
+    "上传json文件X"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import json\n",
+    "data = {\"name\": \"Alice\", \"age\": 20, \"hobbies\": [\"reading\", \"coding\"]}\n",
+    "with open(\"data.json\", \"w\") as f:\n",
+    "    json.dump(data, f)"
+   ]
+  },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "python使用字典操作\n",
+    "\n",
+    "https://www.runoob.com/python/python-dictionary.html"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "<class 'numpy.ndarray'>\n"
+     ]
+    }
+   ],
+   "source": [
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "label, top, left, bottom, right = 'jacket 0.51',538 ,528,707,616\n",
+    "data=(label, top, left, bottom, right)\n",
+    "array = np.asarray(data).reshape(1,5)\n",
+    "array = np.concatenate((array,array),axis=0)\n",
+    "print(type(array))\n",
+    "test=pd.DataFrame(array,columns=['label', 'top', 'left', 'bottom', 'right'])\n",
+    "test.to_csv('save1.csv', index=False, sep=',')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "image_name = 'sadfhkj.jpg'\n",
+    "label, top, left, bottom, right = 'jacket 0.51',538 ,528,707,616\n",
+    "data=(image_name,label, top, left, bottom, right)\n",
+    "array = np.asarray(data).reshape(1,6)\n",
+    "array = np.concatenate((array,array),axis=0)\n",
+    "# image_name = np.asarray(image_name)\n",
+    "# array_image_name = np.concatenate((image_name,array),axis=1)\n",
+    "# label, top, left, bottom, right = array\n",
+    "# array_image_name = image_name + array\n",
+    "\n",
+    "test=pd.DataFrame(array,columns=['image_name','label', 'top', 'left', 'bottom', 'right'])\n",
+    "# test=pd.DataFrame(array)\n",
+    "if os.path.exists('save1.csv'):\n",
+    "    test.to_csv('save1.csv', index=False, sep=',',mode='a', header=None)\n",
+    "    \n",
+    "else:\n",
+    "    test.to_csv('save1.csv', index=False, sep=',',mode='a')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "True"
+      ]
+     },
+     "execution_count": 17,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "import os\n",
+    "os.path.exists('save1.csv')"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "pytorch_py37",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.16"
+  },
+  "orig_nbformat": 4,
+  "vscode": {
+   "interpreter": {
+    "hash": "6198aedcd0ac652a444d7ee87cdb474140afc32862f11a7cd22c6cfd8fc2757e"
+   }
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

get_map.py

@@ -29,7 +29,7 @@
     #   此处的classes_path用于指定需要测量VOC_map的类别
     #   一般情况下与训练和预测所用的classes_path一致即可
     #--------------------------------------------------------------------------------------#
-    classes_path    = 'model_data/voc_classes.txt'
+    classes_path    = 'model_data/cowboy_classes.txt'
     #--------------------------------------------------------------------------------------#
     #   MINOVERLAP用于指定想要获得的mAP0.x，mAP0.x的意义是什么请同学们百度一下。
     #   比如计算mAP0.75，可以设定MINOVERLAP = 0.75。

predict.py

@@ -21,7 +21,7 @@
     #   'export_onnx'       表示将模型导出为onnx，需要pytorch1.7.1以上。
     #   'predict_onnx'      表示利用导出的onnx模型进行预测，相关参数的修改在yolo.py_423行左右处的YOLO_ONNX
     #----------------------------------------------------------------------------------------------------------#
-    mode = "predict"
+    mode = "dir_predict"
     #-------------------------------------------------------------------------#
     #   crop                指定了是否在单张图片预测后对目标进行截取
     #   count               指定了是否进行目标的计数
@@ -158,7 +158,7 @@
             if img_name.lower().endswith(('.bmp', '.dib', '.png', '.jpg', '.jpeg', '.pbm', '.pgm', '.ppm', '.tif', '.tiff')):
                 image_path  = os.path.join(dir_origin_path, img_name)
                 image       = Image.open(image_path)
-                r_image     = yolo.detect_image(image)
+                r_image     = yolo.detect_image(image,image_name=img_name)
                 if not os.path.exists(dir_save_path):
                     os.makedirs(dir_save_path)
                 r_image.save(os.path.join(dir_save_path, img_name.replace(".jpg", ".png")), quality=95, subsampling=0)

train.py

@@ -3,6 +3,7 @@
 #-------------------------------------#
 import datetime
 import os
+os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
 
 import numpy as np
 import torch
@@ -68,7 +69,7 @@
     #   classes_path    指向model_data下的txt，与自己训练的数据集相关 
     #                   训练前一定要修改classes_path，使其对应自己的数据集
     #---------------------------------------------------------------------#
-    classes_path    = 'model_data/voc_classes.txt'
+    classes_path    = 'model_data/cowboy_classes.txt'
     #---------------------------------------------------------------------#
     #   anchors_path    代表先验框对应的txt文件，一般不修改。
     #   anchors_mask    用于帮助代码找到对应的先验框，一般不修改。
@@ -112,7 +113,7 @@
     #                   如果不设置model_path，pretrained = True，此时仅加载主干开始训练。
     #                   如果不设置model_path，pretrained = False，Freeze_Train = Fasle，此时从0开始训练，且没有冻结主干的过程。
     #----------------------------------------------------------------------------------------------------------------------------#
-    pretrained      = False
+    pretrained      = True
     #------------------------------------------------------#
     #   phi             所使用的YoloV5的版本。s、m、l、x
     #                   在除cspdarknet的其它主干中仅影响panet的大小
@@ -187,7 +188,7 @@
     #                           Adam可以使用相对较小的UnFreeze_Epoch
     #   Unfreeze_batch_size     模型在解冻后的batch_size
     #------------------------------------------------------------------#
-    UnFreeze_Epoch      = 300
+    UnFreeze_Epoch      = 100
     Unfreeze_batch_size = 8
     #------------------------------------------------------------------#
     #   Freeze_Train    是否进行冻结训练
@@ -202,7 +203,7 @@
     #   Init_lr         模型的最大学习率
     #   Min_lr          模型的最小学习率，默认为最大学习率的0.01
     #------------------------------------------------------------------#
-    Init_lr             = 1e-2
+    Init_lr             = 1e-3
     Min_lr              = Init_lr * 0.01
     #------------------------------------------------------------------#
     #   optimizer_type  使用到的优化器种类，可选的有adam、sgd
@@ -212,9 +213,9 @@
     #   weight_decay    权值衰减，可防止过拟合
     #                   adam会导致weight_decay错误，使用adam时建议设置为0。
     #------------------------------------------------------------------#
-    optimizer_type      = "sgd"
+    optimizer_type      = "adam"
     momentum            = 0.937
-    weight_decay        = 5e-4
+    weight_decay        = 0
     #------------------------------------------------------------------#
     #   lr_decay_type   使用到的学习率下降方式，可选的有step、cos
     #------------------------------------------------------------------#
@@ -418,6 +419,7 @@
         #   冻结一定部分训练
         #------------------------------------#
         if Freeze_Train:
+            # print("冻结主干")
             for param in model.backbone.parameters():
                 param.requires_grad = False
 

voc_annotation.py

@@ -20,7 +20,7 @@
 #   那么就是因为classes没有设定正确
 #   仅在annotation_mode为0和2的时候有效
 #-------------------------------------------------------------------#
-classes_path        = 'model_data/voc_classes.txt'
+classes_path        = 'model_data/cowboy_classes.txt'
 #--------------------------------------------------------------------------------------------------------------------------------#
 #   trainval_percent用于指定(训练集+验证集)与测试集的比例，默认情况下 (训练集+验证集):测试集 = 9:1
 #   train_percent用于指定(训练集+验证集)中训练集与验证集的比例，默认情况下 训练集:验证集 = 9:1