cl886699

 i'll check the distance loss

maybe it's not trained enough, now i can see some shape of preditions thanks a lot

the training is boken by accident, when i trained enogh , i'll post to tell you the results thanks a lot again

sorry, i edited it again because the bad format i find a bug in my code, when create distance label 1 one hot and get_distance: ``` label[label > 0] =...

hi yes i changed label and then finetuning, i do this because i find the boundary and distance are hard to learn, their losses are much higher than segmentation。and there...

> Does this mean the loss will always be NAN with this code? no it is also can convergence

same problem

FAQ_zh.md文档中有该问题的解决办法 服务启动报错，在api.log中显示：mysql.connector.errors.DatabaseError: 2003 (HY000): Can't connect to MySQL server on 'mysql-container-local:3306' (111) 下载下来后复制导致的问题