Algorithm 1 Decode the output of the detection model

Input:

$The output of dectection model,$

$Tensor=\left\{t_1,\dots ,t_{200}\right\},t_i\in R^6;$

Output:

$All bounding boxes B=\left\{b_1,\dots ,b_{200}\right\},b_i\in R^4;$

$And the corresponding score S=\left\{s_i,\dots ,s_{200}\right\}$, $s_i\in R^1;$

1:  function Decode output$\left(Tensor\right)$

2:  for $i\in \left[0,200\right]$ do

3:    $x_a,y_a,h_a,w_a are priori boxes attr$*/

4:    $\left(x_t,y_t,w_t,h_t,bscore,pscore\right)\leftarrow t_i;$

5:    $x{c}_p =x_t\times w_a\left[i\right]+x_a\left[i\right];$

6:    $y{c}_p=y_t\times \left[i\right]+x_a\left[i\right];$

7:    $h_p=h_a\left[i\right]\times e^{ht};$

8:    $w_p=w_a\left[i\right]\times e^{wt};$

9:    $x_{min}=x{c}_p-w_p/2;$

10:    $x_{max}=x{c}_p+w_p/2;$

11:    $y_{min}=y{c}_p-h_p/2;$

12:    $y_{max}=y{c}_p+h_p/2;$

13:    $b_i=\left(x_{min},y_{min},x_{max},y_{max}\right)$:

14:    $s_i=pscore$

15:    end for

16:  return $B,S$

17:  end function