Classification of Clothing Using Midlevel Layers

Algorithm 1

The process of C-S to determine the final selection mask for each category. CC_𝑤, NC_𝑤, SM_𝑤, and RM_𝑤 are binary vectors that represent if a characteristic is used (1) or removed (0), out of the 27 listed in Table 2, for each category. ≫ stands for a bit shift in the binary vector. Result(⋅) calculates the resulting classification percentage after running the new selection mask on the training set.

Input: Characteristic vector for 𝑤th category (CC_𝑤)

Output: Selection mask for 𝑤th category (SM_𝑤)

SM_𝑤=2²⁶−1;

RM_𝑤=2²⁷−1;

foreach  Article_𝑖 ∈ Category_𝑤  do

 NC_𝑤=(CC_𝑤 × SM_𝑤)∩RM;

 If  Result(NC_𝑤) ≥ Result(CC_𝑤)  then

  RM_𝑤=RM_𝑤 ∩ 2²⁷−2^𝑖−1;

 end

 SM_𝑤=SM_𝑤 ≫ 1;

end