search

qiaoguan / Person-reid-GAN-pytorch

A Pytorch Implementation of "Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro"(ICCV17)
158 stars 49 forks source link

when the baseline result is high(rank-1, 0.934), its difficult to improve the result #1

Open ghost opened 6 years ago

ghost commented 6 years ago

thanks for your sharing,in your conclusion that "when the baseline result is high(rank-1, 0.934), its difficult to improve the result." why occurs this result, does gan not work ?

qiaoguan commented 6 years ago

because this method is very easy, when the accuracy is high, its difficult to improve the result(this applies to any method), if you want to improve the result, you can improve the loss function and you also can try better GAN!