Algorithm 1: Procedure used by the integrated detector
Input	A sliding window $\left(b_x,b_y,b_w,b_h\right)$, face bounding boxes $\left(f_x^n,f_y^n,f_w^n,f_h^n\right),$ face scores $s^n, n=1,2,\dots N$.
Settings	$s_{all}=0$, $m=0$, set $offset$, $r_w$, $r_h$, $f_{ss}$, $d_{hr}$, $f_{wr}$, $f_{hr}$ and $s_{thr}$.
Output	Qualified $\left(b_x,b_y,b_w,b_h\right)$, $s_{all}$.
Step 1	For each $n$: If Equations (2)–(7) are fulfilled:                $m=m+1$          $s_{sc}^m=f_{ss}\cdot \frac{1}{2}\ln \left(\frac{s^n}{1-s^n}\right)\cdot \left(1-\frac{d^n}{d_{hr} \cdot b_h}\right)$             $s_{all}=\max \left\{ s_{sc}^m\right\}, m=1,2,\dots M$
Step 2	Send $s_{all}$ to the cascading decision trees.
Step 3	If $s_{all}=\max \left\{s_{sc}^m\right\}$ is implemented:   If $s_{all}>s_{thr}$:     Output $\left(b_x,b_y,b_w,b_h\right)$; else if $s_{all}>s_{default}$ (default threshold):    Output $\left(b_x,b_y,b_w,b_h\right)$.