Iterating through dictionaries – FutureWarning: Value based partial slicing on non-monotonic DatetimeIndexes with non-existing keys is deprecated

I see you already resolved this issue, but I want to clarify for posterity that the root cause of this warning is the inconsistent spacing in the time series index, while at the same time slicing with keys that do not exist in the index.
There are two main remedies:

1. Use df_motion.resample() and resample to the frequency that’s appropriate (possibly 10ms in this case), providing an aggregation function of some kind, then fill in the blanks as necessary, depending on the change in frequency and the data itself. This will address the "non-monotonic" aspect of the error and allow you to continue slicing with keys that don’t exist in the index.
   df_motion = df_motion.resample(freq='10ms').mean()
2. Slice with keys that exist exactly in the index, e.g. df_motion.loc['2020-07-10 13:49:11.429':'2020-07-10 13:49:11.450']. This will work the same way that loc normally does when dealing with a non-time-series dataframe, and will not raise this error.

As an aside, you’re also creating a duplicate variable in your for loop. When you iterate over a dictionary’s items and assign them temporarily to k, v, k will hold each key, one by one, and v will hold each associated value. So inside of the for loop, you could use v instead of creating the df_motion variable:

audience = {}
for k,v in df_motion_dict.items():
    v = v.loc['2020-07-10 13:55' : '2020-07-10 14:25.5', :]
    if len(v)>0: # remove any dataframes not containing data
        audience[k]=v

or more succinctly:

audience = {
    k: v.loc['2020-07-10 13:55':'2020-07-10 14:25.5']
    for k, v in df_motion_dict.items()
    if not v.empty
}