Vehicle Reidentification Based on MAPANet and k-Reciprocal Encoding

Algorithm 1

Multiattention part feature extraction algorithm.

	Input:
	Output: multiattention part feature vectors ()
(1)	for in //channel grouping
(2)	for in
(3)	//weight vector
(4)	end for
(5)	//weight vector collection
(6)	if
(7)
(8)	else
(9)
(10)	end if
(11)	end for
(12)	for in //generate local attention feature map
(13)	//accumulation of similar channels, using sigmoid to generate probability
(14)	//loss function training
(15)	//features of the jth channel, multiplied element by element
(16)	for in
(17)	//using 1 × 1 convolution to change the number of channels
(18)	//enter the remaining block to obtain the local feature map vector
(19)	end for
(20)
(21)	END