LSong

@YMing-Li Maybe some params need to be changed. Im not sure. As an alternative, you could try https://github.com/open-mmlab/OpenUnReID for your environment.

I forgot. You can add one :simple_smile:

For evaluation, you don't need to care about the last fc layer.

Hi @geyutang , Thanks for your interest in our paper. The M is generated by reranking. Thus, it is not a simple pair-wise distance.

@zengkaiwei We are doing it under in the context of domain adaptation. So we think both datasets matters.

Hi, this reposity uses metric_learn 0.3.0. It this the same with your settings?

Hi @zengkaiwei , we use PK sampling for triplet loss. Please see https://github.com/LcDog/DomainAdaptiveReID/blob/d078d4cc3de951f4e680e613e504664a9befb7a4/reid/utils/data/sampler.py#L11 for more details.

@zengkaiwei 我们follow的是 https://arxiv.org/pdf/1703.07737.pdf 中介绍的方法. 请参考文中3.5的Batch Generation and Augmentation.

@zengkaiwei 可以发邮件联系原作者要

Hi @jinx-USTC , the model is trained with 0.3. You may need to train one by yourself for your specific settings.