【debug】目标检测任务报错TypeError: Argument ‘bb‘ has incorrect type (expected numpy.ndarray, got list)

🚀Debug专栏

        其他debug记录请参考上方【debug专栏】

目录

🚀Debug专栏

🍀🍀背景

❓❓报错

🌻🌻解决方案

1.分析原因

2.增加条件判断

至此就解决该问题了😎😎

3.如果还报错，比如ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (15,) + inhomogeneous part.

4. 报错原因分析

5.继续修改代码

6.修改后完整python代码

至此真的就没有bug了，如果有，留言我们一起debug

---

🍀🍀背景

        调用ann2RLE函数时，有的识别结果像素非常小，没有过滤，直接转RLE格式的txt文本格式，报错TypeError: Argument ‘bb‘ has incorrect type (expected numpy.ndarray, got list)

❓❓报错

        详细报错信息：

        pycocotools 包处调用coco.py 报错TypeError: Argument 'bb' has incorrect type (expected numpy.ndarray, got list）

---

🌻🌻解决方案

1.分析原因

        报这个错误是传入的segment点是四个以内（包含四个）的会触发的错误。

2.增加条件判断

        找到环境安装位置/data/XXX/anaconda3/envs/openmmlab/lib/python3.8/site-packages/pycocotools-2.0-py3.8-linux-x86_64.egg/pycocotools的coco.py文件，修改420行，增加条件判断就好了✔️。

        python代码展示如下：

if len(segm[0])<5:

    segm[0].append(0)

        修改后python完整代码如下：

    def annToRLE(self, ann):
        """
        Convert annotation which can be polygons, uncompressed RLE to RLE.
        :return: binary mask (numpy 2D array)
        """
        t = self.imgs[ann['image_id']]
        h, w = t['height'], t['width']
        segm = ann['segmentation']
        if len(segm[0])<5:           #增加此行
            segm[0].append(0)        #增加此行
        if type(segm) == list:
            # polygon -- a single object might consist of multiple parts
            # we merge all parts into one mask rle code
            rles = maskUtils.frPyObjects(segm, h, w)
            rle = maskUtils.merge(rles)
        elif type(segm['counts']) == list:
            # uncompressed RLE
            rle = maskUtils.frPyObjects(segm, h, w)
        else:
            # rle
            rle = ann['segmentation']
        return rle

至此就解决该问题了😎😎

---

3.如果还报错，比如ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (15,) + inhomogeneous part.

比如我的新报错信息：

  报错信息：File "pycocotools/_mask.pyx", line 303, in pycocotools._mask.frPyObjects
  File "pycocotools/_mask.pyx", line 265, in pycocotools._mask.frPoly
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (15,) + inhomogeneous part.

4. 报错原因分析

        说我的数据是不均匀的，打印了以下segm才发现，是因为我的数据里有长度为3的list，也就是说我的数据中有两个点甚至一个点组成的目标，显然这不是我关心的目标，因为他太小了，所以我索性在这把这种样本删除了。

5.继续修改代码

        增加如下python代码，把各种情况都考虑到。

if len(segm[0])==4:    #如果差一个点，可以补全

    segm[0].append(0)

elif len(segm[0])<4:   #如果差好几个，索性删掉

    segm.remove(segm[0])

6.修改后完整python代码

        完整python代码如下：

def annToRLE(self, ann):
        """
        Convert annotation which can be polygons, uncompressed RLE to RLE.
        :return: binary mask (numpy 2D array)
        """
        t = self.imgs[ann['image_id']]
        h, w = t['height'], t['width']
        segm = ann['segmentation']
#-------------------修改位置----------------------------
        if len(segm[0])==4:    #如果差一个点，可以补全
            segm[0].append(0)
        elif len(segm[0])<4:   #如果差好几个，索性删掉
            segm.remove(segm[0])
#-------------------修改位置----------------------------
        if type(segm) == list:
            # polygon -- a single object might consist of multiple parts
            # we merge all parts into one mask rle code
            rles = maskUtils.frPyObjects(segm, h, w)
            rle = maskUtils.merge(rles)
        elif type(segm['counts']) == list:
            # uncompressed RLE
            rle = maskUtils.frPyObjects(segm, h, w)
        else:
            # rle
            rle = ann['segmentation']
        return rle

至此真的就没有bug了，如果有，留言我们一起debug

---

整理不易，欢迎一键三连！！！

送你们一条美丽的--分割线--

🌷🌷🍀🍀🌾🌾🍓🍓🍂🍂🙋🙋🐸🐸🙋🙋💖💖🍌🍌🔔🔔🍉🍉🍭🍭🍋🍋🍇🍇🏆🏆📸📸⛵⛵⭐⭐🍎🍎👍👍🌷🌷