Zhedong Zheng comments

Results 584 comments of


                                            Zhedong Zheng

Cannot allocate memory || multiprocessing test.py

It looks like the multiprocessing error. You may try to set the `nworkers=1` in DataLoader

how many epochs to reach top performence on market1501?

Hi @bbbxixixixi It is weird. Could you check your Numpy Version first? Is the result correct on the baseline model without PCB?

Use Resnet18 to replace Resnet50

Yes. @NeoYY It is possible. You may check the pretrained model from Pytorch.

TypeError: init() got an unexpected keyword argument 'stride'

Hi @guotong1223 It is wired. Could you show more details? Or did you modify any code?

测试问题test.py

哦哦就是存一下六个关键的东西 query id query feature query camera gallery id gallery feature gallery camera 具体是在这两行 https://github.com/layumi/Person_reID_baseline_pytorch/blob/master/test.py#L300-L301

General inquiry

@ttdd11 , Generally, there are more than one correct images in the gallery (candidate pool). Therefore, I provide the ranking list. Yes. You may consider to add one threshold, or...

General inquiry

@ttdd11 You may add it, however, sometime reranking is time-consuming.

model.py i have meet some problem

Hi @yg76cv Thanks a lot. The default value should be 512 rather than True. I have rectified it. Thanks a lot. @yg76cv

model.py i have meet some problem

Please check my modified code at https://github.com/layumi/Person_reID_baseline_pytorch/blob/master/model.py#L31-L35 Thanks a lot for your feedback.

想把特征维度由2048维降低点，比如，256维或者128维，怎么办

1.现在用的是 512 dim，不是2048。你可以看看怎么学的。 2. 你可以修改模型加一层 256 或者 128在分类层前。