I got them from the article which I linked in my comment.
I read the article. It piled about 5 different assumptions on top of each other.
It takes the upper bound for the deaths in the first 9 months. Then extrapolates it linearly to the last 2.5 years. Then multiplies it by 4, because it estimates that starvation deaths must be 4x.
At that point, why even guess ?
the article is taking projections and applying multiple extrapolations to them. It is a bad estimate.
The current estimates are between 53000 - 66000. About 10x lower than what you propose.