Let's say that you want to detect if a man is running, walking, or dancing instead of just detecting a man still. What type of neural networks will you use for this purpose?
It all depends on the kind of data you have and the type of output you expect.
For example, if you have a series of numbers (heart rate at every minute, or accelerometer data at every second), you can use some time-series approach, like LSTM.
Output / Label
If you want to classify a given segment, you should look for classifiers.
What Bob was doing from 1PM to 2PM?
If you want to find and extract a sub-sample inside a longer period, that's a very different approach.
Find the start and finish time for all running sessions Bob took last week.
In any case, keep in mind the expected prediction format should match the training data format. And some data are more easy to find.