It's self-paced, so it's kinda up to you. There are "deadlines" to help keep you on track, but they're optional. To pass, you just have to pass all the graded assignments by the end date. But if you're getting close to the end and are behind, you can always shift your enrollment to the following session - but you keep all your progress and everything. It's pretty cool in that regard. Very low pressure.
It's been a while since I took it, but I think I spent an hour or two a week watching the videos and reading notes and whatever, and then maybe another 3-5 hours on the programming assignment. It was probably less than that on the earlier programming assignments, and more on the later ones as things got more complicated later on. And I might have spent more time on the videos on certain sections, because of re-watching sections that weren't intuitively clear to me right away. In particular, some of the math'ier stuff where he explained the stuff about using partial derivatives to calculate the error gradients for neural networks... that stuff I had to put more work into since my math background isn't real strong (I never took multi-variable calculus).
All of that said, you can get through the class and learn and understand the material at the level he teaches it, even without completely understanding partial derivatives (a point he makes in the lecture). But having a strong calculus background certainly wouldn't hurt.
I spend about min an hour a day. The general timeline is about 4 hrs per lectures, reading 1 hr reading (mostly optional) and 3 hrs per assignment. I take longer time on assignment since I am new. Sometimes I have to watch lectures over and over again to understand. In general I spend about 10 hrs/week.
I don't think there is any value in paying for the certification, since its an introductory course.
