可视化coco分割标注
import cv2
import random
import json, os
from pycocotools.coco import COCO
from skimage import io
from matplotlib import pyplot as plt
coco_classes = ["111", "112", "113", "121", "122", "123", "131", "132", "133",
"211", "212", "213", "221", "222", "223", "231", "232", "233",
"311", "312", "313", "321", "322", "323", "331", "332", "333",
"411", "412", "413", "421", "422", "423", "431", "432", "433",
"511", "512", "513", "521", "522", "523", "531", "532", "533",
"611", "612", "613", "621", "622", "623", "631", "632", "633",
"711", "712", "713", "721", "722", "723", "731", "732", "733"]
def visualization_bbox_seg(json_path, img_path, *str): # 需要画图的是第num副图片， 对应的json路径和图片路径
coco = COCO(json_path)
if len(str) == 0:
catIds = []
else:
catIds = coco.getCatIds(catNms = [str[0]]) # 获取给定类别对应的id 的dict（单个内嵌字典的类别[{}]）
catIds = coco.loadCats(catIds)[0]['id'] # 获取给定类别对应的id 的dict中的具体id
list_imgIds = coco.getImgIds(catIds=catIds ) # 获取含有该给定类别的所有图片的id
# print(list_imgIds)
for idx in range(len(list_imgIds)):
img = coco.loadImgs(list_imgIds[idx])[0] # 获取满足上述要求，并给定显示第num幅image对应的dict
# print(img)
# # {'license': 1, 'file_name': 'joint_5502.jpg', 'clothes_url': 'None', 'height': 1080, 'width': 1920, 'date_captured': '1998-02-05 05:02:01', 'flickr_url': 'None', 'id': 1}
image = io.imread(os.path.join(img_path, img['file_name'])) # 读取图像
image_name = img['file_name'] # 读取图像名字
image_id = img['id'] # 读取图像id
img_annIds = coco.getAnnIds(imgIds=img['id'], catIds=catIds, iscrowd=None) # 读取这张图片的所有seg_id
# print(img_annIds)
# # [1, 2, 3]
img_anns = coco.loadAnns(img_annIds)
# print(img_anns)
for i in range(len(img_annIds)):
x, y, w, h = img_anns[i-1]['bbox'] # 读取边框
name = coco_classes[img_anns[i-1]['category_id']]
image = cv2.rectangle(image, (int(x), int(y)), (int(x + w), int(y + h)), (0, 255, 255), 1) # 绘制矩形框
cv2.putText(image, name, (int(x+w/2), int(y+h/2)), 5, 3, (255, 0, 0), 3) # 绘制标签
# 各参数依次是：图片，添加的文字，左上角坐标，字体，字体大小，颜色，字体粗细
print(f"{img['file_name']} find {len(img_annIds)} objects")
plt.rcParams['figure.figsize'] = (16.0, 8.5)
plt.imshow(image)
coco.showAnns(img_anns)
plt.show()
# break
if __name__ == "__main__":
train_json = './clothes_val_COCO.json'
train_path = 'hx_clothes_1122/total_val'
visualization_bbox_seg(train_json, train_path) # 最后一个参数不写就是画出一张图中的所有类别
单个json文件合成coco格式大json文件
import sys
import os
import json
from PIL import Image
from tqdm import tqdm
from itertools import chain
START_BOUNDING_BOX_ID = 721 # testing need to change 701 235 721
# If necessary, pre-define category and its id
'''
material：
1、化纤 Chemical-fiber,
2、棉麻 Cotton-linen,
3、羊毛 wool,
4、皮革 leather,
5、皮草 fur,
6、牛仔 jeans,
7、丝质 silk;
color:
1、深色 deep-color,
2、浅色 light-color,
3、白色 white;
style:
1、上衣 jacket,
2、裤子 pants,
3、内衣 underwear
'''
PRE_DEFINE_CATEGORIES = {"111": 1, "112": 2, "113": 3, "121": 4, "122": 5, "123": 6, "131": 7, "132": 8, "133": 9, "211": 10, "212": 11, "213": 12, "221": 13, "222": 14, "223": 15, "231": 16, "232": 17, "233": 18, "311": 19, "312": 20, "313": 21, "321": 22, "322": 23, "323": 24, "331": 25, "332": 26, "333": 27, "411": 28, "412": 29, "413": 30, "421": 31, "422": 32, "423": 33, "431": 34, "432": 35, "433": 36, "511": 37, "512": 38, "513": 39, "521": 40, "522": 41, "523": 42, "531": 43, "532": 44, "533": 45, "611": 46, "612": 47, "613": 48, "621": 49, "622": 50, "623": 51, "631": 52, "632": 53, "633": 54, "711": 55, "712": 56, "713": 57, "721": 58, "722": 59, "723": 60, "731": 61, "732": 62, "733": 63}
# 50000 110971
def convert(jsonsFile, json_file, imgPath):
# ########################################### define the head #################################################
imgs = os.listdir(imgPath)
json_dict = {"info":{}, "licenses":[], "images":[], "annotations": [], "categories": []}
# ######################################### info, type is dict ################################################
info = {'description': 'Clothes Dataset', 'url': 'None', 'version': '1.0', 'year': 2021, 'contributor': 'Donghao Zhangdi etc', 'note': 'material: Chemical-fiber,Cotton-linen,wool,leather,Fur,jeans,silk; color: deep-color,light-color,white; style: jacket,pants,underwear. total 63(7x3x3) categories', 'date_created': '2021/11/25'}
json_dict['info'] = info
# ####################################### licenses, type is list ##############################################
license = {'url': 'None', 'id': 1, 'name': 'None'}
json_dict['licenses'].append(license)
# ###################################### categories, type is list #############################################
categories = PRE_DEFINE_CATEGORIES
for cate, cid in categories.items():
cat = {'supercategory': 'none', 'id': cid , 'name': cate} # no + 1
json_dict['categories'].append(cat)
bnd_id = START_BOUNDING_BOX_ID
jsonnamelist = os.listdir(jsonsFile)
jsonnamelist = [item for item in jsonnamelist if item[-4:] == 'json']
for idx, jsonname in enumerate(tqdm(jsonnamelist)):
# ###################################### images, type is list #############################################
image_id = idx + 236
image_name = jsonname.replace(".json", ".jpg")
if image_name not in imgs:
with open('./error.txt', 'a') as target:
info = f'No image file in image path:n{jsonname} ==> {image_name}nn'
target.write(info)
continue
img = Image.open(os.path.join(imgPath, image_name))
width, height = img.size
image = {'license': 1, 'file_name': image_name, 'clothes_url': 'None', 'height': height, 'width': width, 'date_captured': '1998-02-05 05:02:01', 'flickr_url': 'None', 'id': image_id}
json_dict['images'].append(image)
# ###################################### annotations, type is list #############################################
json_path = os.path.join(jsonsFile, jsonname)
with open(json_path, 'r') as load_f:
load_dict = json.load(load_f)
for obj in load_dict:
label = obj['name']
if label not in categories.keys():
new_id = len(categories)
categories[label] = new_id+1
category_id = categories[label]
points = obj['points']
pointsList = list(chain.from_iterable(points))
pointsList = [float(p) for p in pointsList] ####### point 必须是浮点数！！！！！！！！！！！！！！
seg = [pointsList]
row = pointsList[0::2]
clu = pointsList[1::2]
left_top_x = min(row)
left_top_y = min(clu)
right_bottom_x = max(row)
right_bottom_y = max(clu)
wd = right_bottom_x - left_top_x
hg = right_bottom_y - left_top_y
ann = {'segmentation': seg, 'area': wd*hg, 'iscrowd': 0, 'image_id': image_id, 'bbox': [left_top_x, left_top_y, wd, hg], 'category_id': category_id, 'id': bnd_id}
json_dict['annotations'].append(ann)
bnd_id = bnd_id + 1
print(image_id, bnd_id)
# ######################################### write into local ################################################
with open(json_file, 'w') as json_fp:
json.dump(json_dict, json_fp)
if __name__ == '__main__':
# jsonsFile = "hx_clothes_1122/hx_clothes_masks"
# imgPath = "hx_clothes_1122/hx_clothes_imgs"
# destJson = "./clothes_train_COCO.json"
jsonsFile = "hx_clothes_1122/total_train_jsons"
imgPath = "hx_clothes_1122/total_train"
destJson = "./clothes_train_COCO.json"
# jsonsFile = "hx_clothes_1122/total_val_jsons"
# imgPath = "hx_clothes_1122/total_val"
# destJson = "./clothes_val_COCO.json"
convert(jsonsFile, destJson, imgPath)
