陕西建设机械官方网站,网站建设属于什么支出,免费个人简历模板网站,科技创新与应用从零开始的目标检测和关键点检测#xff08;一#xff09;#xff1a;用labelme标注数据集 1、可视化标注结果2、划分数据集3、Lableme2COCO#xff0c;将json文件转换为MS COCO格式 前言#xff1a;前段时间用到了mmlab的mmdetction和mmpose#xff0c;因此以一个小的数… 从零开始的目标检测和关键点检测一用labelme标注数据集 1、可视化标注结果2、划分数据集3、Lableme2COCO将json文件转换为MS COCO格式 前言前段时间用到了mmlab的mmdetction和mmpose因此以一个小的数据集复现了从数据集制作到模型训练和测试的全流程。希望对想入门mmlab框架的小伙伴有所帮助。主要想做目标检测和关键点检测因此标注目标检测框和关键点。标注范式注意关键点只能在一个目标检测框内
数据链接以标注并转为MS COCO格式数据集 提取码heak 1、可视化标注结果
查看原图片
import cv2
import numpy as np
import json
import matplotlib.pyplot as pltimg_path IMG_4801_JPG_jpg.rf.121fd7907310ecece04dc4020e07e7f2.jpg
img_bgr cv2.imread(img_path)plt.imshow(img_bgr[:,:,::-1])
plt.show()打印json文件信息
labelme_path IMG_4801_JPG_jpg.rf.121fd7907310ecece04dc4020e07e7f2.json
with open(labelme_path, r, encodingutf-8) as f:labelme json.load(f)
print(labelme.keys())
# dict_keys([version, flags, shapes, imagePath, imageData, imageHeight, imageWidth])
print(labelme[shapes])可视化框rectangle标注
# 可视化框rectangle标注
# 框rectangle可视化配置
bbox_color (255, 129, 0) # 框的颜色
bbox_thickness 5 # 框的线宽
# 框类别文字
bbox_labelstr {font_size:2, # 字体大小font_thickness:4, # 字体粗细offset_x:0, # X 方向文字偏移距离向右为正offset_y:-80, # Y 方向文字偏移距离向下为正
}
for each_ann in labelme[shapes]: # 遍历每一个标注if each_ann[shape_type] rectangle: # 筛选出框标注# 框的类别bbox_label each_ann[label]# 框的两点坐标bbox_keypoints each_ann[points]bbox_keypoint_A_xy bbox_keypoints[0]bbox_keypoint_B_xy bbox_keypoints[1]# 左上角坐标bbox_top_left_x int(min(bbox_keypoint_A_xy[0], bbox_keypoint_B_xy[0]))bbox_top_left_y int(min(bbox_keypoint_A_xy[1], bbox_keypoint_B_xy[1]))# 右下角坐标bbox_bottom_right_x int(max(bbox_keypoint_A_xy[0], bbox_keypoint_B_xy[0]))bbox_bottom_right_y int(max(bbox_keypoint_A_xy[1], bbox_keypoint_B_xy[1]))# 画矩形画框img_bgr cv2.rectangle(img_bgr, (bbox_top_left_x, bbox_top_left_y), (bbox_bottom_right_x, bbox_bottom_right_y),bbox_color, bbox_thickness)# 写框类别文字图片文字字符串文字左上角坐标字体字体大小颜色字体粗细img_bgr cv2.putText(img_bgr, bbox_label, (bbox_top_left_x bbox_labelstr[offset_x], bbox_top_left_y bbox_labelstr[offset_y]),cv2.FONT_HERSHEY_SIMPLEX, bbox_labelstr[font_size], bbox_color,bbox_labelstr[font_thickness])
plt.imshow(img_bgr[:,:,::-1])
plt.show()可视化关键点keypoint标注
# 可视化关键点keypoint标注
# 关键点配色
kpt_color_map {head:{id:0, color:[255,0,0], radius:10, thickness:-1},tail:{id:1, color:[0,255,0], radius:10, thickness:-1}
}# 点类别文字
kpt_labelstr {font_size:1, # 字体大小font_thickness:4, # 字体粗细offset_x:30, # X 方向文字偏移距离向右为正offset_y:100, # Y 方向文字偏移距离向下为正
}for each_ann in labelme[shapes]: # 遍历每一个标注if each_ann[shape_type] point: # 筛选出关键点标注kpt_label each_ann[label] # 该点的类别# 该点的 XY 坐标kpt_xy each_ann[points][0]kpt_x, kpt_y int(kpt_xy[0]), int(kpt_xy[1])# 该点的可视化配置kpt_color kpt_color_map[kpt_label][color] # 颜色kpt_radius kpt_color_map[kpt_label][radius] # 半径kpt_thickness kpt_color_map[kpt_label][thickness] # 线宽-1代表填充# 画圆画该关键点img_bgr cv2.circle(img_bgr, (kpt_x, kpt_y), kpt_radius, kpt_color, kpt_thickness)# 写该点类别文字图片文字字符串文字左上角坐标字体字体大小颜色字体粗细img_bgr cv2.putText(img_bgr, kpt_label, (kpt_x kpt_labelstr[offset_x], kpt_y kpt_labelstr[offset_y]),cv2.FONT_HERSHEY_SIMPLEX, kpt_labelstr[font_size], kpt_color,kpt_labelstr[font_thickness])plt.imshow(img_bgr[:,:,::-1])
plt.show()保存可视化图片
cv2.imwrite(visualize.jpg, img_bgr)2、划分数据集
import os
import shutil
import randomfrom tqdm import tqdmDataset_root ../glue_134_Keypoint
os.chdir(os.path.join(Dataset_root, labelme_jsons))print(共有 {} 个 labelme 格式的 json 文件.format(len(os.listdir())))test_frac 0.2 # 测试集比例
random.seed(123) # 随机数种子便于复现folder .img_paths os.listdir(folder)
random.shuffle(img_paths) # 随机打乱val_number int(len(img_paths) * test_frac) # 测试集文件个数
train_files img_paths[val_number:] # 训练集文件名列表
val_files img_paths[:val_number] # 测试集文件名列表print(数据集文件总数, len(img_paths))
print(训练集文件个数, len(train_files))
print(测试集文件个数, len(val_files))# 创建文件夹存放训练集的 labelme格式的 json 标注文件
train_labelme_jsons_folder train_labelme_jsons
os.mkdir(train_labelme_jsons_folder)for each in tqdm(train_files):src_path os.path.join(folder, each)dst_path os.path.join(train_labelme_jsons_folder, each)shutil.move(src_path, dst_path)# 创建文件夹存放训练集的 labelme格式的 json 标注文件
val_labelme_jsons_folder val_labelme_jsons
os.mkdir(val_labelme_jsons_folder)for each in tqdm(val_files):src_path os.path.join(folder, each)dst_path os.path.join(val_labelme_jsons_folder, each)shutil.move(src_path, dst_path)3、Lableme2COCO将json文件转换为MS COCO格式
导入相关包
import os
import json
import numpy as np指定数据集信息
Dataset_root ../glue_134_Keypointclass_list {supercategory: glue,id: 1,name: glue,keypoints: [head, tail], skeleton: [[0, 1]]
}
函数-处理单个labelme标注json文件
def process_single_json(labelme, image_id1):输入labelme的json数据输出coco格式的每个框的关键点标注信息global ANN_IDcoco_annotations []for each_ann in labelme[shapes]: # 遍历该json文件中的所有标注if each_ann[shape_type] rectangle: # 筛选出个体框# 个体框元数据bbox_dict {}bbox_dict[category_id] 1bbox_dict[segmentation] []bbox_dict[iscrowd] 0bbox_dict[segmentation] []bbox_dict[image_id] image_idbbox_dict[id] ANN_ID# print(ANN_ID)ANN_ID 1# 获取个体框坐标bbox_left_top_x min(int(each_ann[points][0][0]), int(each_ann[points][1][0]))bbox_left_top_y min(int(each_ann[points][0][1]), int(each_ann[points][1][1]))bbox_right_bottom_x max(int(each_ann[points][0][0]), int(each_ann[points][1][0]))bbox_right_bottom_y max(int(each_ann[points][0][1]), int(each_ann[points][1][1]))bbox_w bbox_right_bottom_x - bbox_left_top_xbbox_h bbox_right_bottom_y - bbox_left_top_ybbox_dict[bbox] [bbox_left_top_x, bbox_left_top_y, bbox_w, bbox_h] # 左上角x、y、框的w、hbbox_dict[area] bbox_w * bbox_h# 筛选出分割多段线for each_ann in labelme[shapes]: # 遍历所有标注if each_ann[shape_type] polygon: # 筛选出分割多段线标注# 第一个点的坐标first_x each_ann[points][0][0]first_y each_ann[points][0][1]if (first_x bbox_left_top_x) (first_x bbox_right_bottom_x) (first_y bbox_right_bottom_y) (first_y bbox_left_top_y): # 筛选出在该个体框中的关键点bbox_dict[segmentation] list(map(lambda x: list(map(lambda y: round(y, 2), x)), each_ann[points])) # 坐标保留两位小数# bbox_dict[segmentation] each_ann[points]# 筛选出该个体框中的所有关键点bbox_keypoints_dict {}for each_ann in labelme[shapes]: # 遍历所有标注if each_ann[shape_type] point: # 筛选出关键点标注# 关键点横纵坐标x int(each_ann[points][0][0])y int(each_ann[points][0][1])label each_ann[label]if (x bbox_left_top_x) (x bbox_right_bottom_x) (y bbox_right_bottom_y) (y bbox_left_top_y): # 筛选出在该个体框中的关键点bbox_keypoints_dict[label] [x, y]bbox_dict[num_keypoints] len(bbox_keypoints_dict)# print(bbox_keypoints_dict)# 把关键点按照类别顺序排好bbox_dict[keypoints] []for each_class in class_list[keypoints]:if each_class in bbox_keypoints_dict:bbox_dict[keypoints].append(bbox_keypoints_dict[each_class][0])bbox_dict[keypoints].append(bbox_keypoints_dict[each_class][1])bbox_dict[keypoints].append(2) # 2-可见不遮挡 1-遮挡 0-没有点else: # 不存在的点一律为0bbox_dict[keypoints].append(0)bbox_dict[keypoints].append(0)bbox_dict[keypoints].append(0)coco_annotations.append(bbox_dict)return coco_annotations函数-转换当前目录下所有labelme格式的json文件
def process_folder():IMG_ID 0ANN_ID 0# 遍历所有 labelme 格式的 json 文件for labelme_json in os.listdir():if labelme_json.split(.)[-1] json:with open(labelme_json, r, encodingutf-8) as f:labelme json.load(f)## 提取图像元数据img_dict {}img_dict[file_name] labelme[imagePath]img_dict[height] labelme[imageHeight]img_dict[width] labelme[imageWidth]img_dict[id] IMG_IDcoco[images].append(img_dict)## 提取框和关键点信息coco_annotations process_single_json(labelme, image_idIMG_ID)coco[annotations] coco_annotationsIMG_ID 1print(labelme_json, 已处理完毕)else:pass
处理训练集json文件并保存和验证文件是否正常
coco {}coco[categories] []
coco[categories].append(class_list)coco[images] []
coco[annotations] []IMG_ID 0
ANN_ID 0path os.path.join(Dataset_root, labelme_jsons, train_labelme_jsons)
os.chdir(path)process_folder()# 保存coco标注文件
coco_path ../../train_coco.json
with open(coco_path, w) as f:json.dump(coco, f, indent2)os.chdir(../../)os.listdir()from pycocotools.coco import COCOmy_coco COCO(train_coco.json)处理验证集json文件并保存和验证文件是否正常
coco {}coco[categories] []
coco[categories].append(class_list)coco[images] []
coco[annotations] []IMG_ID 0
ANN_ID 0path os.path.join(labelme_jsons, val_labelme_jsons)
os.chdir(path)process_folder()
# 保存coco标注文件
coco_path ../../val_coco.json
with open(coco_path, w) as f:json.dump(coco, f, indent2)os.chdir(../../)my_coco COCO(val_coco.json)至此用于mmdetection框架的MS COCO格式文件转换完成
