COCO数据集的下载以及姿态关键点的数据处理

{ "info": info,"licenses": [license],"images": [image],"annotations": [annotation],} info{ "year": int,"version": str,"description": str,"contributor": str,"url": str,"date_created": datetime,} license{ "id": int,"name": str,} image{ "id": int,"width": int,"height": int,"file_name": str,"license": int,"flickr_url": str,"coco_url": str,"date_captured": datetime,} 

"info":{ "description":"This is stable 1.0 version of the 2014 MS COCO dataset.","url":"http://mscoco.org","version":"1.0","year":2014,"contributor":"Microsoft COCO group","date_created":"2015-01-27 09:11:52.357475" },

{ "license":3,"file_name":"COCO_val2014_000000391895.jpg","coco_url":"http://mscoco.org/images/391895","height":360,"width":640,"date_captured":"2013-11-14 11:18:45","flickr_url":"http://farm9.staticflickr.com/8186/8119368305_4e622c8349_z.jpg","id":391895 },

{ "url":"http://creativecommons.org/licenses/by-nc-sa/2.0/","id":1,"name":"Attribution-NonCommercial-ShareAlike License" },

第三部、认识Object Keypoint 类型的标注格式

1，整体JSON文件格式

比如上图中的person_keypoints_train2017.json、person_keypoints_val2017.json这两个文件就是这种格式。

Object Keypoint这种格式的文件从头至尾按照顺序分为以下段落，看起来和Object Instance一样啊：

{
    "info": info,"licenses": [license],"images": [image],"annotations": [annotation],"categories": [category]
}

是的，你打开这两个文件，虽然内容很多，但从文件开始到结尾按照顺序就是这5段。其中，info、licenses、images这三个结构体/类型 在第一节中已经说了，在不同的JSON文件中这三个类型是一样的，定义是共享的。不共享的是annotation和category这两种结构体，他们在不同类型的JSON文件中是不一样的。

images数组和annotations数组的元素数量是相等的，等于图片的数量。

2，annotations字段

这个类型中的annotation结构体包含了Object Instance中annotation结构体的所有字段，再加上2个额外的字段。

新增的keypoints是一个长度为3*k的数组，其中k是category中keypoints的总数量。每一个keypoint是一个长度为3的数组，第一和第二个元素分别是x和y坐标值，第三个元素是个标志位v，v为0时表示这个关键点没有标注（这种情况下x=y=v=0），v为1时表示这个关键点标注了但是不可见（被遮挡了），v为2时表示这个关键点标注了同时也可见。

num_keypoints表示这个目标上被标注的关键点的数量（v>0），比较小的目标上可能就无法标注关键点。

annotation{ "keypoints": [x1,y1,v1,...],"num_keypoints": int,"id": int,"image_id": int,"category_id": int,"segmentation": RLE or [polygon],"area": float,"bbox": [x,y,width,height],"iscrowd": 0 or 1,} 

从person_keypoints_val2017.json文件中摘出一个annotation的实例如下：

{ "segmentation": [[125.12,539.69,140.94,522.43,100.67,496.54,84.85,469.21,73.35,450.52,104.99,342.65,168.27,290.88,179.78,288,189.84,286.56,191.28,260.67,202.79,240.54,221.48,237.66,248.81,243.42,257.44,256.36,253.12,262.11,275.06,299.15,233.35,329.35,207.46,355.24,206.02,363.87,365.3,210.34,373.93,221.84,226.16,350.92,332.22,234.79,314.97,249.17,271.82,313.89,326.83,227.24,352.72,214.29,357.03,212.85,372.85,208.54,395.87,228.67,414.56,245.93,421.75,266.07,424.63,276.13,437.57,284.76,464.9,286.2,479.28,291.96,489.35,310.65,512.36,549.75,244.49,215.73,546.88,199.91,558.38,204.22,565.57,568.45,184.09,575.64,172.58,578.52,145.26,567.01,117.93,551.19,133.75,532.49]],"num_keypoints": 10,"area": 47803.27955,"iscrowd": 0,"keypoints": [0,142,309,1,177,320,2,191,398,237,317,233,426,306,92,452,123,468,251,469,162,551,2],"image_id": 425226,"bbox": [73.35,206.02,300.58,372.5],"category_id": 1,"id": 183126 },

3，categories字段

最后，对于每一个category结构体，相比Object Instance中的category新增了2个额外的字段，keypoints是一个长度为k的数组，包含了每个关键点的名字；skeleton定义了各个关键点之间的连接性（比如人的左手腕和左肘就是连接的，但是左手腕和右手腕就不是）。目前，COCO的keypoints只标注了person category （分类为人）。

定义如下：

{ "id": int,"supercategory": str,"keypoints": [str],"skeleton": [edge] } 

从person_keypoints_val2017.json文件中摘出一个category的实例如下：

{ "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": [[16,14],[14,12],[17,15],[15,13],[12,[6,[7,7],8],9],[8,10],[9,11],[2,3],[1,2],4],[3,5],[4,6],[5,7]] }

参考文献：https://blog.csdn.net/gqixf/article/details/79280224