Accidentally, I found the following solution is faster that the most popular one:
The only possibility that I can image is that the default parameter of delimiters includes a lot of possible chars to be checked. In this case, only space is possible. To provide this explicitly makes it faster.
class Solution(object): def countSegments(self, s): """ :type s: str :rtype: int """ return len(filter(bool, s.split(' ')))