似乎我需要一些帮助来获得pandas python包的句柄。
我有一张图片列表。每个图像由一组描述符向量描述。这些套装可以有不同的尺寸。我想用pandas存储这些数据。
每个图像都有一组向量，存储为np.ndarray[shape=(N_i, 128), dtype=np.uint8]。
N_i是第i个图像中的矢量数。标记范围（0，N_i）是关于第i个图像的特征索引。
每个图像都由图像ID索引。
因此，每个要素都应该通过图像ID +图像的特征索引进行唯一索引。
在python中，这些对象看起来像：
imageid_list = [1, 2, ...]
vecs_list = [np.ndarray, np.ndarray, ....]
如何将这些信息放入熊猫？我尝试将vecs_list放入一个系列，但它只给了我一个对象列表。我希望熊猫能够更多地了解ndarrays（超出它们是对象的事实）
以下是一些示例代码，可以更好地说明我的问题
print('\n+----------')
print('Get a list of image ids (a stands for annotation)')
aid_list = ibs.get_valid_aids()[0:3]
print('aid_list = ')
print(aid_list)
print('L----------')
#
print('\n+----------')
print('Get the set of vectors from each image/anotation')
vecs_list = ibs.get_annot_desc(aid_list)
print('vecs_list = ')
print(vecs_list)
print('L----------')
#__________
print('\n+----------')
print('Try using just the list of ndarrays to create a hierarchy (doesnt work)')
vecs_series = pd.Series(vecs_list, index=aid_list, name='vecs')
print('vecs_series = ')
print(vecs_series)
print('L----------')
#
print('\n+----------')
print('Try mapping each numpy array in the list to a dataframe')
vecs_dflist = map(pd.DataFrame, vecs_list)
print('vecs_dflist = ')
print(vecs_dflist)
print('L----------')
#__________
print('\n+----------')
print('Try using just the list of dataframes to create a hierarchy (doesnt work)')
vecs_dfseries = pd.Series(vecs_dflist, index=aid_list, name='vecs')
print('vecs_dfseries = ')
print(vecs_dfseries)
print('L----------')
这会产生此输出
+----------
Get a list of image ids (a stands for annotation)
aid_list =
[1, 2, 3]
L----------
+----------
Get the set of vectors from each image/anotation
vecs_list =
[array([[ 19, 0, 0, ..., 106, 4, 0],
[ 58, 0, 0, ..., 26, 4, 1],
[ 10, 40, 55, ..., 9, 27, 54],
...,
[ 78, 0, 0, ..., 7, 5, 8],
[ 40, 24, 2, ..., 3, 0, 5],
[ 59, 7, 5, ..., 70, 33, 15]], dtype=uint8), array([[ 0, 2, 13, ..., 29, 27, 4],
[ 29, 21, 8, ..., 11, 5, 7],
[ 1, 1, 2, ..., 0, 4, 3],
...,
[ 10, 27, 39, ..., 35, 85, 23],
[ 1, 27, 115, ..., 88, 2, 1],
[ 31, 1, 2, ..., 15, 10, 5]], dtype=uint8), array([[ 0, 0, 0, ..., 0, 0, 14],
[ 0, 0, 1, ..., 3, 27, 127],
[ 16, 8, 18, ..., 25, 123, 23],
...,
[ 5, 52, 6, ..., 21, 87, 31],
[ 27, 55, 30, ..., 12, 56, 13],
[ 79, 29, 0, ..., 18, 21, 29]], dtype=uint8)]
L----------
+----------
Try using just the list of ndarrays to create a hierarchy (doesnt work)
vecs_series =
1 [[19, 0, 0, 1, 3, 0, 0, 36, 2, 0, 0, 12, 117, ...
2 [[0, 2, 13, 1, 3, 37, 64, 0, 3, 18, 29, 1, 4, ...
3 [[0, 0, 0, 0, 7, 33, 30, 2, 12, 0, 0, 2, 7, 30...
Name: vecs, dtype: object
L----------
+----------
Try mapping each numpy array in the list to a dataframe
vecs_dflist =
[ 0 1 2 3 4 ... 123 124 125 126 127
0 19 0 0 1 3 ... 0 12 106 4 0
1 58 0 0 2 2 ... 11 38 26 4 1
2 10 40 55 37 10 ... 0 0 9 27 54
3 71 6 0 0 1 ... 1 3 0 0 0
4 0 0 3 2 82 ... 4 3 0 0 26
... ... ... ... ... ... ... ... ... ... ... ...
1070 8 14 6 11 14 ... 29 46 15 7 15
1071 23 17 3 19 48 ... 4 7 5 13 52
1072 78 0 0 0 0 ... 11 56 7 5 8
1073 40 24 2 12 42 ... 25 19 3 0 5
1074 59 7 5 2 0 ... 1 21 70 33 15
[1075 rows x 128 columns], 0 1 2 3 4 ... 123 124 125 126 127
0 0 2 13 1 3 ... 0 8 29 27 4
1 29 21 8 2 14 ... 26 23 11 5 7
2 1 1 2 2 3 ... 117 15 0 4 3
3 20 78 43 27 27 ... 20 35 13 25 16
4 8 10 4 0 20 ... 103 23 0 0 0
... ... ... ... ... ... ... ... ... ... ... ...
1203 40 6 0 22 57 ... 13 21 13 26 13
1204 62 9 6 8 13 ... 3 9 10 3 4
1205 10 27 39 13 3 ... 6 11 35 85 23
1206 1 27 115 7 2 ... 16 81 88 2 1
1207 31 1 2 1 5 ... 2 4 15 10 5
[1208 rows x 128 columns], 0 1 2 3 4 ... 123 124 125 126 127
0 0 0 0 0 7 ... 118 1 0 0 14
1 0 0 1 3 6 ... 0 0 3 27 127
2 16 8 18 1 0 ... 0 0 25 123 23
3 2 0 0 1 4 ... 0 0 0 29 94
4 0 12 8 2 0 ... 11 7 3 3 3
.. ... ... ... ... ... ... ... ... ... ... ...
904 0 3 0 0 72 ... 0 0 0 2 14
905 41 11 1 24 46 ... 0 7 21 71 63
906 5 52 6 0 17 ... 0 0 21 87 31
907 27 55 30 5 5 ... 2 0 12 56 13
908 79 29 0 0 100 ... 11 81 18 21 29
[909 rows x 128 columns]]
L----------
+----------
Try using just the list of dataframes to create a hierarchy (doesnt work)
vecs_dfseries =
1 0 1 2 3 4 ... 123 124 ...
2 0 1 2 3 4 ... 123 124 ...
3 0 1 2 3 4 ... 123 124 1...
Name: vecs, dtype: object
L----------