object_id time_id class x y
0 3db53411-c23b-49ec-8635-adc4e3ee2895 5G21A6P01L4100029:1570754223950071 NaN NaN NaN
1 3cea3cdc-883e-48d7-83de-e485da2e085a 5G21A6P01L4100029:1570754223950071 PERSON 528.868 2191.747
2 fc87a12f-a76a-4273-a712-6f56afc042c6 5G21A6P01L4100029:1570754223950071 CAR 512.238 2192.744
3 4edb4e32-0345-4f85-a4b1-e60903368fed 5G21A6S09K40039EX:1565470602550590 NaN NaN NaN
4 cd68a1d0-2470-4096-adb1-201017aadc9e 5G21A6S09K40039EX:1565470602550590 PERSON -1305.968 -2423.231
我有一个嵌套字典 detections
具有以下架构
detections = defaultdict(dict)
detections[key:time_id][key:object_id] = {'class_text':... , 'x': ..., 'y': ...}
对于上述数据框,检测
将是:
detections[5G21A6P01L4100029:1570754223950071] =
{
`3db53411-c23b-49ec-8635-adc4e3ee2895`: {},
'3cea3cdc-883e-48d7-83de-e485da2e085a': {'class_text': 'PERSON', 'x': 528.8, 'y': 2191.7},
'fc87a12f-a76a-4273-a712-6f56afc042c6': {'class_text': 'CAR', 'x': 512.2, 'y': 2192.7}}
}
detections["5G21A6S09K40039EX:1565470602550590"] =
{
`4edb4e32-0345-4f85-a4b1-e60903368fed`: {},
'cd68a1d0-2470-4096-adb1-201017aadc9e': {'class_text': 'PERSON', 'x': -1305.968, 'y': -2423.23}
}
当 (class
, x
and y
) 的值为 NaN 时,detections
为空值,否则它具有相应的值。
对于如何在不对每一行进行循环的情况下进行检测
的任何评论,我都很感激?
最佳答案
在 time_id
上使用 groupby
并应用自定义合并函数 merge_dicts
根据预定义的要求将分组数据帧合并到字典中:
def merge_dicts(s):
s = s.set_index('object_id')[['class', 'x', 'y']]
return s.agg(lambda x: {} if x.isna().all() else dict(**x), axis=1).to_dict()
detections = df.groupby('time_id').apply(merge_dicts).to_dict()
结果:
print(detections)
{
'5G21A6P01L4100029: 1570754223950071':
{
'3db53411-c23b-49ec-8635-adc4e3ee2895': {},
'3cea3cdc-883e-48d7-83de-e485da2e085a': {'class': 'PERSON', 'x': 528.868, 'y': 2191.7470000000003},
'fc87a12f-a76a-4273-a712-6f56afc042c6': {'class': 'CAR', 'x': 512.238, 'y': 2192.744}
},
'5G21A6S09K40039EX: 1565470602550590':
{
'4edb4e32-0345-4f85-a4b1-e60903368fed': {},
'cd68a1d0-2470-4096-adb1-201017aadc9e': {'class': 'PERSON', 'x': -1305.968, 'y': -2423.231}
}
}
关于python - 将具有 NaN 的多列 Pandas 数据框转换为嵌套字典,我们在Stack Overflow上找到一个类似的问题: https://stackoverflow.com/questions/63166521/