imglab .xml 标签格式转coco .json格式

Posted 2021-02-17 jiajiewu

用imglab做物体特征点标注，将标注数据转为coco keypoints数据集转换代码：

import xml.dom.minidom as xmldom
import os
import csv
import json

image_w = 640
image_h = 640
image_start_index = 0
instance_start_index  = 10000

info = {"description": "COCO 2017 Dataset","url": "http://cocodataset.org","version": "1.0","year": 2017,"contributor": "COCO Consortium","date_created": "2017/09/01"}
licenses = [{"url": "http://creativecommons.org/licenses/by-nc-sa/2.0/","id": 1,"name": "Attribution-NonCommercial-ShareAlike License"},
{"url": "http://creativecommons.org/licenses/by-nc/2.0/","id": 2,"name": "Attribution-NonCommercial License"},
{"url": "http://creativecommons.org/licenses/by-nc-nd/2.0/","id": 3,"name": "Attribution-NonCommercial-NoDerivs License"},
{"url": "http://creativecommons.org/licenses/by/2.0/","id": 4,"name": "Attribution License"},
{"url": "http://creativecommons.org/licenses/by-sa/2.0/","id": 5,"name": "Attribution-ShareAlike License"},
{"url": "http://creativecommons.org/licenses/by-nd/2.0/","id": 6,"name": "Attribution-NoDerivs License"},
{"url": "http://flickr.com/commons/usage/","id": 7,"name": "No known copyright restrictions"},
{"url": "http://www.usa.gov/copyright.shtml","id": 8,"name": "United States Government Work"}]
images = []
annotations = []
categories = [{"supercategory": "person","id": 1,"name": "person","keypoints": ["nose","left_eye","right_eye","left_ear","right_ear","left_shoulder","right_shoulder","left_elbow","right_elbow","left_wrist","right_wrist","left_hip","right_hip","left_knee","right_knee","left_ankle","right_ankle"],
"skeleton": []}]

 
xml_filepath = os.path.abspath("/home/jiajie/pytorch/detectron2/datasets/imglab_data/mydata.xml")

############################################
# 得到文件对象
dom_obj = xmldom.parse(xml_filepath)
 
# 得到元素对象
element_obj = dom_obj.documentElement
 
# 获得子标签
 
sub_element_obj = element_obj.getElementsByTagName("image")
 
for i in range(len(sub_element_obj)):
    image_start_index = image_start_index + 1
    #images
    image_info = {"license": 1} 
    image_info[‘file_name‘] = str(sub_element_obj[i].getAttribute("file")).split("\")[0]+"_"+str(sub_element_obj[i].getAttribute("file")).split("\")[1]
    #print(image_info[‘file_name‘])
    image_info[‘coco_url‘] = ""
    image_info[‘height‘] = int(image_h)
    image_info[‘width‘] = int(image_w)
    image_info[‘date_captured‘] = ""
    image_info[‘flickr_url‘] = ""
    image_info[‘id‘] = int(image_start_index)
    images.append(image_info) 
    sub_element_obj_box = sub_element_obj[i].getElementsByTagName("box")
    for j in range(len(sub_element_obj_box)):
        instance_start_index = instance_start_index  + 1
        sub_element_obj_keypoints = sub_element_obj_box[j].getElementsByTagName("part")
        # annotations
        annotations_info = {"segmentation":[[]]}
        annotations_info[‘num_keypoints‘] = int(len(sub_element_obj_keypoints))
        annotations_info[‘area‘] = 0.0
        annotations_info[‘iscrowd‘] = 0
        keypoints_dict = {}
        
        for k in range(len(sub_element_obj_keypoints)):
            #print(sub_element_obj_keypoints[k].getAttribute("x"),sub_element_obj_keypoints[k].getAttribute("y"))
            kp_name = str(sub_element_obj_keypoints[k].getAttribute("name"))+"_"+str(sub_element_obj_keypoints[k].getAttribute("x"))+"_"+str(sub_element_obj_keypoints[k].getAttribute("y"))
            keypoints_dict[kp_name] = pow(int(sub_element_obj_keypoints[k].getAttribute("x")),2)+pow(int(sub_element_obj_keypoints[k].getAttribute("y")),2)            
        sort_list = sorted(keypoints_dict.items(),  key=lambda d: d[1], reverse=False)
        #print(sort_list)
        keypoints_list = []  
        #print(int(str(sort_list[0][0]).split("_")[1]))   
        keypoints_list.append(int(str(sort_list[0][0]).split("_")[1]))
        
        keypoints_list.append(int(str(sort_list[0][0]).split("_")[2]))
        keypoints_list.append(int(2))
        keypoints_list.append(int(str(sort_list[1][0]).split("_")[1]))
        keypoints_list.append(int(str(sort_list[1][0]).split("_")[2]))
        keypoints_list.append(int(2))
        keypoints_list.append(int(str(sort_list[2][0]).split("_")[1]))
        keypoints_list.append(int(str(sort_list[2][0]).split("_")[2]))
        keypoints_list.append(int(2))
        keypoints_list.append(int(str(sort_list[3][0]).split("_")[1]))
        keypoints_list.append(int(str(sort_list[3][0]).split("_")[2]))
        keypoints_list.append(int(2) )
        annotations_info[‘keypoints‘] = keypoints_list
        annotations_info[‘image_id‘] = int(image_start_index)
        #print(keypoints_list)
        #bbox
        annotations_info[‘bbox‘] = [int(sub_element_obj_box[j].getAttribute("left")),int(sub_element_obj_box[j].getAttribute("top")),int(sub_element_obj_box[j].getAttribute("width")),int(sub_element_obj_box[j].getAttribute("height"))]
        annotations_info[‘category_id‘] = 1
        annotations_info[‘id‘] = instance_start_index
        annotations.append(annotations_info)

image = {
    ‘info‘: info,
    ‘licenses‘: licenses,
    ‘images‘: images,
    ‘annotations‘: annotations,
    ‘categories‘: categories
}


with open(‘/home/jiajie/pytorch/detectron2/datasets/imglab_data/person_keypoints_train2017.json‘, ‘w‘) as f:
    json.dump(image,f)