I've been stuck in this problem for a very long time, but today I got the idea of Binary Indexed Tree from wikipedia and finally solved this problem :)
Binary Indexed Tree,or BIT, can be viewed as a simplified version of Segment Tree,easier to code and less function to perform. But It's enough to solve this problem

#include<iostream>
#include<cstdio>
using namespace std;

const int Max = 32005;

int n;
int c[Max] = {0};
int level[Max] = {0};

int lowbit(int x)
{
    return x&(-x);
}

int sum(int i)
{
    int s = 0;
    while(i > 0)
    {
        s += c[i];
        i -= lowbit(i);
    }
    return s;
}

void update(int pos,int val)
{
    while(pos <= Max)
    {
        c[pos] += val;
        pos += lowbit(pos);
    }
}

int main()
{
    scanf("%d",&n);
    for(int i = 0; i < n; i ++)
    {
        int x,y;
        scanf("%d%d",&x,&y);
        x ++;
        level[ sum(x) ] ++;
        update(x,1);
    }

    for(int i = 0; i < n; i ++)    printf("%d\n",level[i]);
}

How can this even be correct if you don't use the y-coordinates at all?

Because in statement there was written that y-coordinates are already sorted

in function main,why you use "x ++"?

cose in problem x in [0,32000 ]

so 0 is bad value for binary :)

so we just shift all x on some const ( just 1) so all value between 1 and power(2,someN)

Why do you use 32005 as a size of an array?

Because it is given in constraint. That x and y values will be less than 32000.