mask rcnn 将mask 转json数据01
2024-04-26 03:06:33
目的：存在 图片及其对应的mask 文件 ，通过Python代码转化为json 数据，无需手工制作
1、首先准备好 图片和及其对应的mask ，使用的mask为黑底白色
#!/usr/bin/env python3
#把 mask和原图 集合到一个json 文件中
import datetime
import json
import os
import re
import fnmatch
from PIL import Image
import numpy as np
from pycococreatortools import pycococreatortoolsROOT_DIR = 'train'
IMAGE_DIR = os.path.join(ROOT_DIR, "shapes_train2018")
ANNOTATION_DIR = os.path.join(ROOT_DIR, "annotations")INFO = {"description": "Example Dataset","url": "https://github.com/waspinator/pycococreator","version": "0.1.0","year": 2018,"contributor": "waspinator","date_created": datetime.datetime.utcnow().isoformat(' ')
}LICENSES = [{"id": 1,"name": "Attribution-NonCommercial-ShareAlike License","url": "http://creativecommons.org/licenses/by-nc-sa/2.0/"}
]
#适当修改类别
CATEGORIES = [{'id': 1,'name': 'square','supercategory': 'shape',},{'id': 2,'name': 'circle','supercategory': 'shape',},{'id': 3,'name': 'triangle','supercategory': 'shape',},
]#返回rgb文件路径 不做修改
def filter_for_jpeg(root, files):file_types = ['*.jpeg', '*.jpg']file_types = r'|'.join([fnmatch.translate(x) for x in file_types])files = [os.path.join(root, f) for f in files]files = [f for f in files if re.match(file_types, f)]return files
#返回mask文件路径 不做修改
def filter_for_annotations(root, files, image_filename):file_types = ['*.png']file_types = r'|'.join([fnmatch.translate(x) for x in file_types])basename_no_extension = os.path.splitext(os.path.basename(image_filename))[0]file_name_prefix = basename_no_extension + '.*'files = [os.path.join(root, f) for f in files]files = [f for f in files if re.match(file_types, f)]files = [f for f in files if re.match(file_name_prefix, os.path.splitext(os.path.basename(f))[0])]return filesdef main():coco_output = {"info": INFO,"licenses": LICENSES,"categories": CATEGORIES,"images": [],"annotations": []}image_id = 1segmentation_id = 1 # filter for jpeg imagesfor root, _, files in os.walk(IMAGE_DIR):image_files = filter_for_jpeg(root, files)# go through each imagefor image_filename in image_files:image = Image.open(image_filename)image_info = pycococreatortools.create_image_info(image_id, os.path.basename(image_filename), image.size)coco_output["images"].append(image_info)# filter for associated png annotationsfor root, _, files in os.walk(ANNOTATION_DIR):annotation_files = filter_for_annotations(root, files, image_filename)# go through each associated annotationfor annotation_filename in annotation_files:print(annotation_filename)class_id = [x['id'] for x in CATEGORIES if x['name'] in annotation_filename][0]category_info = {'id': class_id, 'is_crowd': 'crowd' in image_filename}binary_mask = np.asarray(Image.open(annotation_filename).convert('1')).astype(np.uint8)annotation_info = pycococreatortools.create_annotation_info(segmentation_id, image_id, category_info, binary_mask,image.size, tolerance=2)if annotation_info is not None:coco_output["annotations"].append(annotation_info)segmentation_id = segmentation_id + 1image_id = image_id + 1with open('{}/instances_shape_train2018.json'.format(ROOT_DIR), 'w') as output_json_file:json.dump(coco_output, output_json_file)if __name__ == "__main__":main()
上面代码是 多个标注数据 数据集合到一个json文件中的
{"info": {"description": "Example Dataset", "url": "https://github.com/waspinator/pycococreator", "version": "0.1.0", "year": 2018, "contributor": "waspinator", "date_created": "2021-03-13 06:54:48.550064"}, "licenses": [{"id": 1, "name": "Attribution-NonCommercial-ShareAlike License", "url": "http://creativecommons.org/licenses/by-nc-sa/2.0/"}], "categories": [{"id": 1, "name": "square", "supercategory": "shape"}], "images": [{"id": 1, "file_name": "0.jpg", "width": 256, "height": 384, "date_captured": "2021-03-13 04:34:17.084304", "license": 1, "coco_url": "", "flickr_url": ""}, {"id": 2, "file_name": "1.jpg", "width": 256, "height": 384, "date_captured": "2021-03-13 04:34:17.084304", "license": 1, "coco_url": "", "flickr_url": ""}, {"id": 3, "file_name": "2.jpg", "width": 384, "height": 256, "date_captured": "2021-03-13 04:34:17.084304", "license": 1, "coco_url": "", "flickr_url": ""}], "annotations": [{"id": 1, "image_id": 1, "category_id": 1, "iscrowd": 0, "area": 130, "bbox": [106.0, 109.0, 31.0, 15.0], "segmentation": [[126.0, 123.5, 113.0, 121.5, 105.5, 109.0, 118.0, 120.5, 136.5, 115.0, 126.0, 123.5]], "width": 256, "height": 384}, {"id": 2, "image_id": 2, "category_id": 1, "iscrowd": 0, "area": 1734, "bbox": [51.0, 158.0, 41.0, 69.0], "segmentation": [[74.0, 226.5, 54.5, 213.0, 50.5, 194.0, 70.5, 160.0, 76.0, 157.5, 81.5, 161.0, 82.5, 171.0, 91.5, 177.0, 91.5, 185.0, 84.5, 210.0, 74.0, 226.5]], "width": 256, "height": 384}, {"id": 3, "image_id": 3, "category_id": 1, "iscrowd": 0, "area": 2444, "bbox": [233.0, 127.0, 33.0, 104.0], "segmentation": [[254.0, 230.5, 237.5, 229.0, 244.5, 220.0, 242.5, 191.0, 232.5, 175.0, 236.5, 171.0, 232.5, 166.0, 235.5, 150.0, 245.0, 126.5, 254.5, 130.0, 264.5, 149.0, 264.5, 224.0, 254.0, 230.5]], "width": 384, "height": 256}]}2 下面的代码是 ，每对数据 对应一个json 文件（只能实现mask中 只有一个类的情况）
#!/usr/bin/env python3
#功能批量将mask 转单个json import datetime
import json
import os
import io
import re
import fnmatch
import json
from PIL import Image
import numpy as np
from pycococreatortools import pycococreatortools
from PIL import Image
import base64
from base64 import b64encodeROOT_DIR = 'train1'
IMAGE_DIR = os.path.join(ROOT_DIR, "pic")
ANNOTATION_DIR = os.path.join(ROOT_DIR, "annotations")def img_tobyte(img_pil):
# 类型转换 重要代码# img_pil = Image.fromarray(roi)ENCODING='utf-8'img_byte=io.BytesIO()img_pil.save(img_byte,format='PNG')binary_str2=img_byte.getvalue()imageData = base64.b64encode(binary_str2)base64_string = imageData.decode(ENCODING)return base64_stringannotation_files=os.listdir(ANNOTATION_DIR)
for annotation_filename in annotation_files:coco_output = {"version": "3.16.7","flags": {},"fillColor": [255, 0,0,128],"lineColor": [0,255,0, 128],"imagePath": {},"shapes": [],"imageData": {},"imageHeight": 512,"imageWidth": 512}print(annotation_filename)class_id = 1name = annotation_filename.split('.',3)[0]name1=name+'.jpg'coco_output["imagePath"]=name1 image = Image.open(IMAGE_DIR+'/'+ name1)imageData=img_tobyte(image)coco_output["imageData"]= imageData binary_mask = np.asarray(Image.open(ANNOTATION_DIR+'/'+annotation_filename).convert('1')).astype(np.uint8)segmentation=pycococreatortools.binary_mask_to_polygon(binary_mask, tolerance=3)list1=[]for i in range(0, len(segmentation[0]), 2):list1.append( [segmentation[0][i],segmentation[0][i+1]])seg_info = {'points': list1, "fill_color":'null' ,"line_color":'null' ,"label": "1", "shape_type": "polygon","flags": {}}coco_output["shapes"].append(seg_info)full_path='{}/'+name+'.json'with open( full_path.format(ROOT_DIR), 'w') as output_json_file:json.dump(coco_output, output_json_file)3 需要自己改写成 自己需要的文件
#!/usr/bin/env python3
#功能批量将多个同类mask 转单个json #筛选多余的点集合 可能mask制作的不规范，产生多余的点，这样的话，利用点的数量，做一个限制
#我这里设置的是10 可根据具体效果改动for item in segmentation:if(len(item)>10):list1=[]for i in range(0, len(item), 2):list1.append( [item[i],item[i+1]])seg_info = {'points': list1, "fill_color":'null' ,"line_color":'null' ,"label": "1", "shape_type": "polygon","flags": {}}coco_output["shapes"].append(seg_info)coco_output[ "imageHeight"]=binary_mask.shape[0]coco_output[ "imageWidth"]=binary_mask.shape[1]full_path='{}/'+name+'.json'with open( full_path.format(ROOT_DIR), 'w') as output_json_file:json.dump(coco_output, output_json_file)完善的代码
#!/usr/bin/env python3
#功能批量将多个同类mask 转单个json import datetime
import json
import os
import io
import re
import fnmatch
import json
from PIL import Image
import numpy as np
from pycococreatortools import pycococreatortools
from PIL import Image
import base64
from base64 import b64encodeROOT_DIR = 'C:/Users/11549/Desktop/pycococreator-master/examples/shapes/train2/pic/'
IMAGE_DIR = os.path.join(ROOT_DIR, "pic")
ANNOTATION_DIR = os.path.join(ROOT_DIR, "mask")def img_tobyte(img_pil):
# 类型转换 重要代码# img_pil = Image.fromarray(roi)ENCODING='utf-8'img_byte=io.BytesIO()img_pil.save(img_byte,format='PNG')binary_str2=img_byte.getvalue()imageData = base64.b64encode(binary_str2)base64_string = imageData.decode(ENCODING)return base64_stringannotation_files=os.listdir(ANNOTATION_DIR)
for annotation_filename in annotation_files:coco_output = {"version": "3.16.7","flags": {},"fillColor": [255, 0,0,128],"lineColor": [0,255,0, 128],"imagePath": {},"shapes": [],"imageData": {} }print(annotation_filename)class_id = 1name = annotation_filename.split('.',3)[0]name1=name+'.jpg'coco_output["imagePath"]=name1 image = Image.open(IMAGE_DIR+'/'+ name1)imageData=img_tobyte(image)coco_output["imageData"]= imageData binary_mask = np.asarray(Image.open(ANNOTATION_DIR+'/'+annotation_filename).convert('1')).astype(np.uint8)segmentation=pycococreatortools.binary_mask_to_polygon(binary_mask, tolerance=3)#筛选多余的点集合for item in segmentation:if(len(item)>10):list1=[]for i in range(0, len(item), 2):list1.append( [item[i],item[i+1]])seg_info = {'points': list1, "fill_color":'null' ,"line_color":'null' ,"label": "1", "shape_type": "polygon","flags": {}}coco_output["shapes"].append(seg_info)coco_output[ "imageHeight"]=binary_mask.shape[0]coco_output[ "imageWidth"]=binary_mask.shape[1]full_path='{}/'+name+'.json'with open( full_path.format(ROOT_DIR), 'w') as output_json_file:json.dump(coco_output, output_json_file)4 生成的json文件，效果如下
{"version": "3.20.0","flags": {},"shapes": [{"label": "1","line_color": null,"fill_color": null,"points": [[362.59375,166.1875],[357.90625,216.96875],[429.78125,216.96875],[436.03125,171.65625]],"shape_type": "polygon","flags": {}}],"lineColor": [0,255,0,128],"fillColor": [255,0,0,128],"imagePath": "3.jpg","imageData": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL/2wBDAQkJCQwLDBgNDRgyIRwhMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjIyMjL/wAARCAIAAgADASIAAhEBAxEB/8QAHwAAAQUBAQEBAQEAAAAAAAAAAAECAwQFBgcICQoL/8QAtRAAAgEDAwIEAwUFBAQAAAF9AQIDAAQRBRIhMUEGE1FhByJxFDKBkaEII0KxwRVS0fAkM2JyggkKFhcYGRolJicoKSo0NTY3ODk6Q0RFRkdISUpTVFVWV1hZWmNkZWZnaGlqc3R1dnd4eXqDhIW QS+trmElPmVv057GpFfU//9k=","imageHeight": 512,"imageWidth": 512
}这样我们就有了 json数据和图片数据了，可以安装步骤生成 coco格式的数据l
5 把json数据和图片数据放在一起，利用各bat脚本 生成labelme_json 数据 如下图所示。
@echo off
for %%i in (*.json) do labelme_json_to_dataset "%%i"
pause
6 然而 其中的label.png 文件，并不可用 需要转为8位的数据 才可以使用
下面是批量的代码
import numpy as np
from PIL import Imageimport os
YourFile_path='labelme_json/'
files = os.listdir(YourFile_path)
path1='cv2_mask/'for item in files:src=YourFile_path+item+'/label.png'img= Image.open(src)img = Image.fromarray(np.uint8(img))name = item.split("_")[0]name=name+'.png'img.save( path1+name)最终生成 可以使用的数据 输入mask -rcnn中，
7 零碎的代码
#针对一个box的检测程序，输出x , y, w, h 文件名
#统计bbox的程序
import cv2
import os
import operator
path='mask/'def takeSecond1(elem): return list(elem)[2]*list(elem)[3]
list1=[ ]
filelist = os.listdir(path)
for item in filelist:bgr_img = cv2.imread(path+item)gray_img = cv2.cvtColor(bgr_img, cv2.COLOR_BGR2GRAY)gray_img = cv2.blur(gray_img, (5,5))th, binary = cv2.threshold(gray_img, 125, 255, cv2.THRESH_OTSU)contours, hierarchy = cv2.findContours(binary, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)bounding_boxes = [cv2.boundingRect(cnt) for cnt in contours]bounding_boxes.sort(key=takeSecond1)list1.append([item,bgr_img.shape,bounding_boxes[-1]])# cv2.drawContours(bgr_img, contours, -1, (0, 0, 255), 3)# for i in range(len(bounding_boxes)):# print(i)# cv2.imshow("name", bgr_img)# cv2.waitKey(6000)# cv2.destroyAllWindows()#coco 可视化代码
from pycocotools.coco import COCO
import numpy as np
import skimage.io as io
import matplotlib.pyplot as plt
import pylabimage_directory = 'train/shapes_train2018/'
annotation_file = 'C:/Users/11549/Desktop/pycococreator-master/examples/shapes/train/instances_shape_train2018.json'
example_coco = COCO(annotation_file)categories = example_coco.loadCats(example_coco.getCatIds())
category_names = [category['name'] for category in categories]
print('Custom COCO categories: \n{}\n'.format(' '.join(category_names)))category_names = set([category['supercategory'] for category in categories])
print('Custom COCO supercategories: \n{}'.format(' '.join(category_names)))category_ids = example_coco.getCatIds(catNms=['square'])
image_ids = example_coco.getImgIds(catIds=category_ids)
image_data = example_coco.loadImgs(image_ids[np.random.randint(0, len(image_ids))])[0]image_data# load and display instance annotations
image = io.imread(image_directory + image_data['file_name'])
plt.imshow(image); plt.axis('off')
pylab.rcParams['figure.figsize'] = (8.0, 10.0)
annotation_ids = example_coco.getAnnIds(imgIds=image_data['id'], catIds=category_ids, iscrowd=None)
annotations = example_coco.loadAnns(annotation_ids)
example_coco.showAnns(annotations)#重命名代码
import os
path = 'C:/Users/11549/Desktop/pycococreator-master/examples/shapes/train1/shapes_train2018/cv2mask/'
filelist = os.listdir(path)
for i in range(len(filelist)):print(i)src = os.path.join(os.path.abspath(path),filelist[i])dst = os.path.join(os.path.abspath(path),str(i) + '.png')os.rename(src,dst)import os
import cv2
path = 'pic/'
path1 = '1/'
# os.access(path+'0002-01.jpg', os.F_OK)
filelist = os.listdir(path)
for item in filelist:src = os.path.join(os.path.abspath(path),item)img = cv2.imread(src) res=cv2.resize(img,(757,568))cv2.imwrite(path1+item, res)8 相应的安装包 下载地址：https://download.csdn.net/download/weixin_44576543/15806642
last one 由于时间比较急 ，所以比较粗糙 大家不懂的可以留言评论
mask rcnn 将mask 转json数据01相关推荐
- two-stage-anchor-based-faster-rcnn进阶-mask rcnn：Mask R-CNN
paper:https://arxiv.org/abs/1703.06870 code:https://github.com/matterport/Mask_RCNN mask rcnn是基于fast ...
- 完整代码+实操！手把手教你操作Faster R-CNN和Mask R-CNN
点击上方↑↑↑蓝字关注我们~ 「2019 Python开发者日」全日程揭晓,请扫码咨询 ↑↑↑ 机器视觉领域的核心问题之一就是目标检测(Object Detection),它的任务是找出图像当中所有感 ...
- 目标分割：Mask RCNN
日萌社 人工智能AI:Keras PyTorch MXNet TensorFlow PaddlePaddle 深度学习实战(不定时更新) 目标分割:Mask RCNN 目标分割:Mask RCNN 气 ...
- Mask Rcnn训练自己的航拍数据集
最近想利用Mask Rcnn训练自己的航拍数据(数据来源:自己采集.飞行高度:200m,图片大小WxH=6000x4000),在进行自己的航拍图片训练时也与其他同学进行交流,他们一般进行识别的是一类物 ...
- 图像目标分割_6 Mask RCNN
6.6.0 背景 目标检测和语义分割的效果在短时间内得到了很大的改善.在很大程度上,这些进步是由强大的基线系统驱动的,例如,分别用于目标检测和语义分割的Fast/Faster R-CNN和全卷积网络( ...
- 卷积神经网络——实例分割之Mask R-CNN论文翻译
论文链接 https://arxiv.org/abs/1703.06870 相关论文翻译链接 R-CNN:https://blog.csdn.net/itlilyer/article/details/ ...
- Mask R-CNN论文
目录 摘要 1.引言 2.相关工作 3.Mask R-CNN 3.1 实施细节 4.实验:实例分割 4.1 主要结果 4.2 消融实验 4.3 边界框检测结果 4.4 时序 5.Mask R-CNN ...
- Mask R-CNN全文翻译
文章目录 标题:Mask R-CNN 资源下载(PDF.学习笔记及资料) Abstract 1.Introduction:(机翻) 2.Related Work:(机翻) 3.Mask R-CNN 3 ...
- Mask R-CNN 原理解析
Kaiming He Georgia Gkioxari Piotr Doll´ ar Ross Girshick Facebook AI Research (FAIR) 摘要 我们提出了一个概念上简单 ...
最新文章
- 上海世博会信息化的8大看点
- Binder Driver浅析：Binder线程池
- 5中排序算法（冒泡，选择，插入，快速，归并）
- 2017-2018-1 JAVA实验站 第三周作业
- EF Core For MySql查询中使用DateTime.Now作为查询条件的一个小问题
- 28和lba48命令格式区别_编译Sass（命令行）
- 企业微信怎么删除好友 企业微信如何删除成员
- vue 导入excel解析_Vue实现Excel导入并解析
- 2014-VGG网络讲解
- Android时代的赢创之路
- 学习python: 2.x 和 3.x的区别
- python变量存为matlab,将matlab变量导出为python用法的文本
- WPF中的TreeView入门
- Chrome浏览器的翻译插件开发
- 超详细“图+文”讲解，带你快速入门STM32编程！
- 腾讯云内容分发网络 CDN 产品认证课程笔记(二)——腾讯云CDN介绍
- oracle卸数的perl脚本,卸载cpan安装的所有perl模块
- APP从苹果开发者A账号转移到B账号的流程
- MySQL编程：将查询到的字段赋值给变量
- 制作各种docker镜像