arcface-pytorch The loss function does not converge

The loss function does not converge

Open hyperion0819 opened this issue 3 years ago • 7 comments

use my own datasets,however,the acc of train.py always equals 0.At the same time,the loss function does not converge.Why? i need adjust hyper-parameter(s or m)?

Apr 27 '22 11:04 hyperion0819

No, it needs lots of datasets. How many images do you have?

Apr 27 '22 15:04 bubbliiiing

No, it needs lots of datasets. How many images do you have?

您好我使用的是鱼体数据集有30个类别 358张图片但我将这些图片进行了复制得到了500个文件夹 5000张图片（这些图片与原图片没有区别）这样可以吗

Apr 28 '22 01:04 hyperion0819

复制？？复制为什么可以到500哥文件夹？那同一个类在不同文件夹不就错了吗

Apr 28 '22 02:04 bubbliiiing

复制？？复制为什么可以到500哥文件夹？那同一个类在不同文件夹不就错了吗

嗯嗯明白了总之就是需要大量数据集其次不用更改s m参数训练的准确率就能提升是吗

Apr 28 '22 02:04 hyperion0819

s m参数是啥……我其实都不知道

Apr 28 '22 02:04 bubbliiiing

head里面的么？不需要改

Apr 28 '22 02:04 bubbliiiing

head里面的么？不需要改

好的谢谢

Apr 28 '22 02:04 hyperion0819

arcface-pytorch arcface-pytorch copied to clipboard

The loss function does not converge

arcface-pytorch
arcface-pytorch copied to clipboard