【pycocotools】评价个人数据集：如何输出更多不同IoU的AP值

最新推荐文章于 2025-03-26 15:23:11 发布

下雨天不下雨

最新推荐文章于 2025-03-26 15:23:11 发布

阅读量539

点赞数 5

文章标签：目标检测深度学习 python

本文链接：https://blog.youkuaiyun.com/retainenergy/article/details/130173690

版权

修改IoU，输出多种不同IoU的AP值

代码定位，找到 pycocotools/cocoeval.py ，463行，修改为以下情况：

        def _summarizeDets():
            stats = np.zeros((31,))
            stats[0] = _summarize(1)
            stats[1] = _summarize(1, iouThr=.5, maxDets=self.params.maxDets[2])
            stats[2] = _summarize(1, iouThr=.75, maxDets=self.params.maxDets[2])
            stats[3] = _summarize(1, areaRng='small', maxDets=self.params.maxDets[2])
            stats[4] = _summarize(1, areaRng='medium', maxDets=self.params.maxDets[2])
            stats[5] = _summarize(1, areaRng='large', maxDets=self.params.maxDets[2])
            stats[6] = _summarize(0, maxDets=self.params.maxDets[0])
            stats[7] = _summarize(0, maxDets=self.params.maxDets[1])
            stats[8] = _summarize(0, maxDets=self.params.maxDets[2])
            stats[9] = _summarize(0, areaRng='small', maxDets=self.params.maxDets[2])
            stats[10] = _summarize(0, areaRng='medium', maxDets=self.params.maxDets[2])
            stats[11] = _summarize(0, areaRng='large', maxDets=self.params.maxDets[2])

            stats[12] = _summarize(1, iouThr=0.6, maxDets=self.params.maxDets[2])
            stats[13] = _summarize(1,  iouThr=0.7,maxDets=self.params.maxDets[2])
            stats[14] = _summarize(1,  iouThr=0.8,maxDets=self.params.maxDets[2])
            stats[15] = _summarize(1,  iouThr=0.85,maxDets=self.params.maxDets[2])

            stats[16] = _summarize(1,  areaRng='small', iouThr=0.5,maxDets=self.params.maxDets[2])
            stats[17] = _summarize(1,  areaRng='small', iouThr=0.6,maxDets=self.params.maxDets[2])
            stats[18] = _summarize(1,  areaRng='small', iouThr=0.7,maxDets=self.params.maxDets[2])
            stats[19] = _summarize(1,  areaRng='small', iouThr=0.8,maxDets=self.params.maxDets[2])
            stats[20] = _summarize(1,  areaRng='small', iouThr=0.85,maxDets=self.params.maxDets[2])

            stats[21] = _summarize(1,  areaRng='medium', iouThr=0.5,maxDets=self.params.maxDets[2])
            stats[22] = _summarize(1, areaRng='medium', iouThr=0.6,maxDets=self.params.maxDets[2])
            stats[23] = _summarize(1,  areaRng='medium', iouThr=0.7,maxDets=self.params.maxDets[2])
            stats[24] = _summarize(1, areaRng='medium', iouThr=0.8,maxDets=self.params.maxDets[2])
            stats[25] = _summarize(1,  areaRng='medium', iouThr=0.85,maxDets=self.params.maxDets[2])

            stats[26] = _summarize(1,  areaRng='large', iouThr=0.5,maxDets=self.params.maxDets[2])
            stats[27] = _summarize(1, areaRng='large', iouThr=0.6,maxDets=self.params.maxDets[2])

            stats[28] = _summarize(1, areaRng='large', iouThr=0.7,maxDets=self.params.maxDets[2])
            stats[29] = _summarize(1, areaRng='large', iouThr=0.8,maxDets=self.params.maxDets[2])
            stats[30] = _summarize(1, areaRng='large', iouThr=0.85,maxDets=self.params.maxDets[2])

            return stats

更改后输出效果：

 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.60      | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.70      | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.80      | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.89      | area=   all | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50      | area= small | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.60      | area= small | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.70      | area= small | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.80      | area= small | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50      | area=medium | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.60      | area=medium | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.70      | area=medium | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.80      | area=medium | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.89      | area=medium | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.50      | area= large | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.60      | area= large | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.70      | area= large | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.80      | area= large | maxDets=100 ] = 0
 Average Precision  (AP) @[ IoU=0.89      | area= large | maxDets=100 ] = 0

修改COCO评价指标 maxDets=[10,15,20]

该指标的意思是分别保留测试集的每张图上置信度排名第1、前10、前100个预测框，根据这些预测框和真实框进行比对，来计算AP、AR等值。
我需要的 maxDets=[10,15,20]
但我用的是自己数据集，只是数据格式是coco的格式，评价标准还按原本的就不合适了
举个例子：如果你测试集一张图上面的目标超过100个，你还用 maxDets=[1,10,100]，这结果肯定有问题了

根据我自己的数据集，测试集的图片上的目标不超过20个，我需要的 maxDets=[10,15,20]，要达到这个目的只要添加一行代码即可：

找到coco_evaluation.py，在代码末尾找到coco_eval，然后参照下图进行添加一行代码即可

'''添加在此处'''
coco_eval.params.maxDets=[10,15,20]##
coco_eval.evaluate()
coco_eval.accumulate()
coco_eval.summarlize()

return coco_eval

参考链接

1、https://blog.youkuaiyun.com/weixin_45798949/article/details/127752343?ops_request_misc=&request_id=&biz_id=102&utm_term=pycocotools%E8%AF%84%E4%BB%B7%E6%95%B0%E6%8D%AE%E9%9B%86%E9%BB%98%E8%AE%A4%E7%9A%84iou%E8%AE%BE%E7%BD%AE&utm_medium=distribute.pc_search_result.none-task-blog-2_allsobaiduweb~default-0-127752343.142^v83wechat,239^v2insert_chatgpt&spm=1018.2226.3001.4187

2、https://blog.youkuaiyun.com/weixin_42899627/article/details/120689553
3、coco和yolov5 map计算结果不一致的问题