the recognition result did not rise, but decreased

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

the recognition result did not rise, but decreased #22

Open XXXxiaoxiaoxia opened 5 years ago

XXXxiaoxiaoxia commented 5 years ago

Why did the recognition accuracy drop when I added the image generated by gan? At the same time, the recognition accuracy did not reach the accuracy in the paper.