Xinchen Liu
Xinchen Liu
I have the same question. Hope to meet the authors at ICCV19.
> I guess the problem is due to the version of conda libs such as pytorch, cuda, etc., because the version of libs used by our codebase may be too...
> 您好,在engine/trainer.py中我发现from solver.build import make_lr_scheduler, make_optimizer。我从这个reid-strong-baseline找到了solver,但是它仅仅有make_optimizer,我是一个新手,很抱歉打扰你,您能告诉我从哪里获得make_lr_scheduler吗? > 另一个问题就是我阅读了您的论文,这个parse后的数据集我也准备好了,我想问PE(part erasing augmentation)的实现在哪里,看起来似乎只有RE,如果我想仅仅训练Global Branch + PE,我仅仅需要使用Veri_mask或者vehicleid_mask数据集吗,在哪一步进行了PE? 期待您的回复,很抱歉打扰你,谢谢!! PE的实现代码在这个地方:https://github.com/lxc86739795/vehicle_reid_by_parsing/blob/master/modeling/baseline_selfgcn.py 训练的代码在这里: https://github.com/lxc86739795/vehicle_reid_by_parsing/blob/master/tools/train_selfgcn.py solver目录之前进行了修改,现在已经更新。