Thanks to the author for his contribution. The results I reproduced with the released code and the results of the paper are quite different, some indicators are poor, and some indicators are better. Is that reasonable?
MODE: test || mAP ==> tool: 0.86692 || verb: 0.64725 || target: 0.5572 || triplet: 0.38674 || iv: 0.37391 || it: 0.4392 ----
Thanks to the author for his contribution. The results I reproduced with the released code and the results of the paper are quite different, some indicators are poor, and some indicators are better. Is that reasonable?
MODE: test || mAP ==> tool: 0.86692 || verb: 0.64725 || target: 0.5572 || triplet: 0.38674 || iv: 0.37391 || it: 0.4392 ----