Yixiao Ge

> 您好，之前跟您讨论过计算jacard距时矩阵过大的问题。您跟我提过分chunk的思路。我现在的想法是先用个mini-batch k-means去粗分成几个chunk，您觉得这样的思路可以吗。或者您觉得有什么更好的分chunk的思路吗。麻烦您啦! 我没有想好具体的，我一开始想的就是简单地把数据集分为N个chunk，然后分别做聚类。但是这样的话，不同chunk之间的数据无法连通，会损失部分精度。需要考虑如何在获得聚类结果的情况下将不同的chunk进行整合（聚类合并），我记得有一些传统方法做了这件事情，类似于并行聚类的算法，你可以去具体查阅一下。

> 还想问下您，分chunk后每个chunk聚类得到多个indep_thres。对于多个indep_thres您觉得怎么处理会好一些呢。 我的意思是分chunk进行聚类，但是获得完聚类label后还是整合在一起，所以还会只是一组indep_thres。

Sure, I will add them after training. But I was busy these days, you could try to train on your own if you need the models urgently. Thanks.

Actually you only need to modify the dataset name and the image size to ``` DATA: height: 224 width: 224 ``` Could you please provide your training config and the...

Thanks for the suggestion. I will consider it in the next version.

@zwzhang121 Please help solve it.

@Reggieharden @muzishen Hi, actually the VehicleID dataset does not provide camera IDs, as well as the query and gallery split. Thus conventional metrics (e.g. mAP, CMC) do not support the...

Hi, it is under development. We are a little busy these days, and we will try to support SSG ASAP. Thanks.

A good question. Actually, I am not sure about it. I used such an identity sampler according to its wide usage in person re-id applications. You could try to modify...

Hi Zhiwei, thanks for your pulling requests. I have made some comments, please check and modify the requests accordingly.