supyer9

Results 4 issues of supyer9

![85981141-48b3a100-ba16-11ea-9a29-e7d0413ad43d](https://user-images.githubusercontent.com/54097420/86114268-e5506e80-bafc-11ea-80a6-4b98214d5005.png) ![85981178-536e3600-ba16-11ea-8810-cd4c0eeb0b62](https://user-images.githubusercontent.com/54097420/86114293-ea152280-bafc-11ea-9282-43ea4d80bfa4.png) ![85981196-59fcad80-ba16-11ea-8952-7e25c3bef9d8](https://user-images.githubusercontent.com/54097420/86114297-eb464f80-bafc-11ea-8cdf-5fe9acd44bac.png) 您好,不好意思打扰了。我修改了一下代码,使代码能在每跑完一个epoch后画出验证集的曲线。但是我发现验证集的曲线有一些问题。heatmap的验证集曲线没有下降反而直接增加了。我训练集验证集的比例为8:1。是因为验证集的数据分布和训练集太像了,所以还没有训练完一个epoch就发生了过拟合?我不确定到底是什么问题,您能帮忙解答一下吗?万分感谢。

您好, 感谢您对MOT任务做出的贡献。我是一个MOT新手,阅读了FairMot源码的issues发现您在下方进行了回复,将源码改成了多类别跟踪。 我想请教下,除了更改update函数以外,还需要更改哪些内容呢?在FairMot源码的基础上修改可以吗? 谢谢,祝好

@bernhard2202 @d-acharya @zzhiwu I am very interested in your work, but when I run the program, there is a bug. The bug is located at utils.py->def convert_img(). I want to...

![image](https://user-images.githubusercontent.com/54097420/85981141-48b3a100-ba16-11ea-9a29-e7d0413ad43d.png) ![image](https://user-images.githubusercontent.com/54097420/85981178-536e3600-ba16-11ea-8810-cd4c0eeb0b62.png) ![image](https://user-images.githubusercontent.com/54097420/85981196-59fcad80-ba16-11ea-8952-7e25c3bef9d8.png) ![image](https://user-images.githubusercontent.com/54097420/85981208-61bc5200-ba16-11ea-9bf0-b4a00a02a7f4.png) Hello!I modified your code so that it can verify the validation set after training an epoch training set. Although the loss curve of the training set...