hh23333 / tsd Goto Github PK

Dear Author,

I have read your paper in detail and am greatly interested in your research findings. At the same time, I must extend my sincere congratulations to you for the significant progress you have made in the performance of occluded person re-identification models.

I have reviewed your publicly available code and attempted to reproduce your model. However, I regret to inform you that I have not been able to achieve similar results. Therefore, I would very much appreciate your assistance and guidance. If you could offer some suggestions on the configuration files and training steps, I would be extremely grateful.

您好，我想请问一下，您的mask generator那一部分，您使用全连接层进行分类从而划分人体部分，可见性为什么要用预测呢，这里我不太懂，训练的可见性标签又是什么呀，因为一般都是直接根据概率判断的，谢谢啦

作者大大，可以分享一下可视化热图文件（vis_attn.py）如何调用吗？感激不尽~