Xia Meng
Xia Meng
i meet the same problem, is it still not solved?
I found the reason. After the training, the model ckpt will be stored in several profiles. However, the most importment one, which has the suffix 'ckpt', is default invisible. The...
> I suggest you try to save the checkpoints at different times, like 1k epoch 10k epoch 20k epoch, and check whether the model weights are abnormal. Thanks~ I did....
same problem +1
I met the same problem, i tested step by step and finally found the wrong place. It is an absolute path problem. The lfw_root in config.py shouldn't begin with '/data/...',...
2022-09-02 09:06:45,228 - mmseg - INFO - Iter [650/100000] lr: 9.947e-04, eta: 1:44:39, time: 0.063, data_time: 0.002, memory: 1485, decode.loss_ce: 0.0028, decode.acc_seg: 93.3247, loss: 0.0028 2022-09-02 09:06:48,408 - mmseg -...
Part of the output is above. After several thousands of iter it is still the same. I thought maybe it is the config problem. I started from the example config...
您好，我已收到您的邮件，我将尽快查阅并给予回复，谢谢！（孟夏的自动回复）
@xiexinch Thank you for your reply~ I will do some check. I have tried both using the pretrained model and not using it, the only difference is the decode.acc_seg and...
where is the tools/detect.py? I can only see generate_seg_tusimple.py