datatrucks转coco格式

怼烎@ 2022-09-07 14:57 192阅读 0赞

import json
    import os
    
    import cv2
    import numpy as np
    
    # with open('/media/lixuan/Data/MSCOCO/annotations_trainval2017/annotations/instances_train2017.json') as f:
    #     datas = json.load(f)
    #     """
    #     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:[{'license': 2, 'file_name': '000000015335.jpg', 'coco_url': 'http://images.cocodataset.org/val2017/000000015335.jpg', 'height': 480, 'width': 640, 'date_captured': '2013-11-25 14:00:10', 'flickr_url': 'http://farm6.staticflickr.com/5533/10257288534_c916fafd78_z.jpg', 'id': 15335}]
    #     annotations:[{'segmentation': {'counts': [169679, 8, 465, 17, 462, 19, 460, 21, 458, 23, 457, 23, 457, 23, 457, 23, 457, 23, 457, 23, 457, 23, 458, 22, 459, 20, 461, 18, 464, 6, 4, 3, 51836, 9, 470, 22, 457, 25, 454, 27, 453, 27, 453, 28, 452, 28, 452, 28, 452, 28, 453, 1, 4, 21, 461, 19, 461, 18, 461, 18, 462, 16, 10071, 12, 468, 13, 467, 14, 464, 1, 1, 15, 465, 15, 462, 2, 1, 15, 462, 18, 462, 18, 462, 18, 462, 18, 471, 8, 472, 7, 448, 4, 21, 4, 452, 4, 471, 1, 4, 5, 469, 12, 468, 12, 468, 12, 468, 12, 468, 12, 468, 12, 468, 12, 469, 10, 471, 8, 474, 3, 51123], 'size': [480, 640]}, 'area': 964, 'iscrowd': 1, 'image_id': 562818, 'bbox': [353, 229, 180, 39], 'category_id': 1, 'id': 900100562818}]
    #     categories[{'supercategory': 'indoor', 'id': 90, 'name': 'toothbrush'}]
    #
    #     'area': 702.1057499999998, 'iscrowd': 0, 'image_id': 289343, 'bbox': [473.07, 395.93, 38.65, 28.67], 'category_id': 18, 'id': 1768
    #     segmentation
    #     area
    #     iscrowd
    #     image_id
    #     bbox
    #     category_id
    #     id
    #
    #     """
    
    def getbox(cnt):
        cnt = np.array(cnt)
        rect = cv2.minAreaRect(cnt) # 得到最小外接矩形的（中心(x,y), (宽,高), 旋转角度）
        box = cv2.boxPoints(rect) # 获取最小外接矩形的4个顶点坐标(ps: cv2.boxPoints(rect) for OpenCV 3.x)
        return [int(np.min(box[:,0])),int(np.min(box[:,1])),int(np.max(box[:,0])),int(np.max(box[:,1]))]
    
    def getarea(points):
        return cv2.contourArea(np.array(points))
    
    def convertcoco(pathlist):
        CocoDIct = {}
        CocoDIct['info'] = {'description': 'COCO 2017 Dataset', 'url': 'http://cocodataset.org', 'version': '1.0', 'year': 2017, 'contributor': 'COCO Consortium', 'date_created': '2017/09/01'}
        CocoDIct['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'}]
        CocoDIct['images'] = []
        CocoDIct['annotations'] = []
        CocoDIct['categories'] = [{'supercategory': 'icecream', 'id': 1, 'name': 'clear'},{'supercategory': 'icecream', 'id': 2, 'name': 'not_so_clear'}]
        image_id = 1
        id = 1
        for path in pathlist:
            with open(path) as f:#/home/lixuan/workspace/dataset/icecreamtwo/6000
                datas = f.readlines()
                for data in datas:
                    image = {}
                    data = data.strip()
                    dict = json.loads(data)
                    image['license'] = 1
                    image['file_name'] = dict['content'].split('/')[-1]
                    image['coco_url'] = dict['content']
                    try:
                        for anotation in dict['annotation']:
                            annotation = {}
                            segmentation = {}
                            label = anotation['label']
                            points = anotation['points']
                            imageWidth = anotation['imageWidth']
                            imageHeight = anotation['imageHeight']
                            counts = []
                            if [None,None] not in points:
                                for point in points:
                                    counts.append(point[0] * imageWidth)
                                    counts.append(point[1] * imageHeight)
                                    point[0] = int(point[0] * imageWidth)
                                    point[1] = int(point[1] * imageHeight)
                                area = getarea(points)
                                bbox = getbox(points)
                                segmentation['counts'] = counts
                                segmentation['size'] = [imageHeight,imageWidth]
                                annotation['segmentation'] = segmentation
                                annotation['area'] = area
                                annotation['iscrowd'] = 1
                                annotation['image_id'] = image_id
                                annotation['bbox'] = bbox
                                annotation['category_id'] = 1 if label == ['clear'] else 2
                                annotation['id'] = id
                                id += 1
                            if annotation != {}:
                                CocoDIct['annotations'].append(annotation)
                        image['height'] = imageHeight
                        image['width'] = imageWidth
                        image['date_captured'] = '2013-11-25 14:00:10'
                        image['flickr_url'] = dict['content']
                        image['id'] = image_id
                        image_id += 1
                        CocoDIct['images'].append(image)
                    except:
                        print(dict['annotation'])
        with open('train.json','w') as f:
            json.dump(CocoDIct,f)
    
    bsdir1 = '/home/lixuan/workspace/dataset/icecreamtwo'
    convertcoco(['IcecreamSeg_G1.json'])#,os.path.join(bsdir1,'iceSeg_dxy42279.json'),os.path.join(bsdir1,'iceSeg_nick.json'),os.path.join(bsdir1,'iceSeg_韦浩萍.json'),
                 # os.path.join(bsdir1,'iceSeg_panda.json'),os.path.join(bsdir1,'iceSeg_taku.json'),os.path.join(bsdir1,'iceSeg_xichen_LINQING.json'),
                 # os.path.join(bsdir1,'iceSeg_zhenglixue_Luo.json'),os.path.join(bsdir1,'iceSeg_楚丹.json'),os.path.join(bsdir1,'iceSeg_郑燕静.json')])