Algorithm 1 Overall algorithm with the proposed method

Input: An image sequence from a depth information video Output: An image sequence where touching-pigs are individually separated // Load a depth information video $Seq=\mathrm{Load}\left(\mathrm{depth}\_\mathrm{video}\right)$ // Remove noises in depth images $Se{q}_{Interpolate}=\mathrm{SpatioTemporalInterpolate}\left(Seq\right)$ // Perform background subtraction $Se{q}_{BS}=\mathrm{BackgroundSubtract}\left(Se{q}_{Interpolate}\right)$ // Detect moving pigs only $Se{q}_{GMM}=\mathrm{GMM}\left(Se{q}_{BS}\right)$ // Separate touching-pigs using the proposed method $\mathrm{for}\left(i=1;i\le \mathrm{the}\mathrm{number}\mathrm{of}\mathrm{connected}\mathrm{components}\mathrm{in}Se{q}_{GMM};i++\right)$   If size of each connected component $I_{in}^i\left(Se{q}_{GMM}\right)$ $>$ size of a single pig     Go to YOLO Processing Module with $I_{in}^i\left(Se{q}_{GMM}\right)$     Save the separated results of the touching-pigs Return;