I have the following dataset with recurring classes that I would like to attach individual IDs to. The challenge is that these classes are dynamic, ranging from 0 to x (depending on how many objects of the same class I detect).
If I detect 5 people at a certain time, I will get/attach the classes [0, 1, 2, 3, 4] If I get 2 people at a later point of time, I will get/attach the classes [0, 1]
Whenever I get a new class, I would like to attach an individual ID to the individual classes, so that the first 4 classes keep their initial class label [0, 1, 2, 3], while the later classes follow with [4, 5], instead of [0, 1].
See the uniqueID for the desired output.
The data looks like this:
class datetime uniqueID
0 2023-11-24T15:45 1
1 2023-11-24T15:45 2
2 2023-11-24T15:45 3
3 2023-11-24T15:45 4
4 2023-11-24T15:45 5
0 2023-11-24T15:46 1
1 2023-11-24T15:46 2
2 2023-11-24T15:46 3
3 2023-11-24T15:46 4
4 2023-11-24T15:46 5
nan 2023-11-24T15:47 nan
0 2023-11-24T15:48 6
1 2023-11-24T15:48 7
0 2023-11-24T15:49 6
1 2023-11-24T15:49 7
0 2023-11-24T15:50 6
1 2023-11-24T15:50 7
2 2023-11-24T15:50 8
3 2023-11-24T15:50 9
4 2023-11-24T15:50 10
0 2023-11-24T15:51 6
1 2023-11-24T15:51 7
nan 2023-11-24T15:52 nan
0 2023-11-24T15:53 11
1 2023-11-24T15:53 12
2 2023-11-24T15:53 13
3 2023-11-24T15:53 14
4 2023-11-24T15:53 15
0 2023-11-24T15:54 11
1 2023-11-24T15:54 12
2 2023-11-24T15:54 13
3 2023-11-24T15:54 14
0 2023-11-24T15:55 11
1 2023-11-24T15:56 12
2 2023-11-24T15:57 13
3 2023-11-24T15:58 14
nan 2023-11-24T15:59 nan
nan 2023-11-24T16:00 nan
0 2023-11-24T16:01 15
0 2023-11-24T16:02 15
Whenever there is a nan, the classes are new and should be attached with a new unique ID. I tried to append the classes to a list, and work with the datetime to determine whether the unique classes are "new" or existing.
Any help how to achieve this?