Skip to main content
. 2022 Sep 27;8:e1098. doi: 10.7717/peerj-cs.1098

Table 7. Performance for different supervision strategies shown in Fig. 5.

Index Train Test MSMT17
Strategy Final feature mAP (%) Rank-1 (%)
(a) Single-loss Add 64.1 82.4
(b) Single-loss Concat 64.6 83.0
(c) Single-loss Multiply 56.6 75.4
(d1) Duplicate-loss Add 64.2 81.8
(d2) Duplicate-loss Concat 66.9 84.5