Deep Person Re-identification

[AlexeyAB]

Bag of Tricks and A Strong Baseline for Deep Person Re-identification

• paper: https://arxiv.org/abs/1903.07071v3

• code: https://github.com/michuanhaohao/reid-strong-baseline

• comparison, 2nd place at 05.03.2020: https://paperswithcode.com/sota/person-re-identification-on-market-1501

• cosine distance:

  • https://en.wikipedia.org/wiki/Cosine_similarity
  • https://cmry.github.io/notes/euclidean-v-cosine
  • pHash, Hamming distance - for preliminary search in a large database

double cosine_similarity(double *A, double *B, unsigned int size)
{
    double mul = 0.0, d_a = 0.0, d_b = 0.0 ;

    for(unsigned int i = 0; i < size; ++i) 
    {
        mul += A[i] * B[i] ;
        d_a += A[i] * A[i] ;
        d_b += B[i] * B[i] ;
    }
    double similarity = 0;
    double divider = sqrt(d_a) * sqrt(d_b);
    if(divider > 0) similarity = mul / divider;
    return similarity;
}

---

---

---

[WongKinYiu]

add some references:

1. Input Two Continuous Images, Estimate Offset of Center of Objects Tracking Objects as Points https://github.com/xingyizhou/CenterTrack

2. Online Single Stage Joint Detection and Tracking RetinaTrack: Online Single Stage Joint Detection and Tracking