ShiMinghao0208

Results 7 issues of ShiMinghao0208

How can I get the pretrained weight? I want to add some structure to vit,then finetune and observe the result

DATA: data_root: dataset/voc2012 train_list: dataset/voc2012/list/train_aug.txt val_list: dataset/voc2012/list/val.txt i couldn't find these txt files ,i download ade20k and voc2012. And i don't know how to put the dataset dir. ![image](https://user-images.githubusercontent.com/53176521/97657363-a4fdfe80-1aa4-11eb-90f2-f6d16d82ab03.png)

About matrix,the l1 norm is caculate the max column norm. So do we think of FS as a matrix? How to understand the l1 norm here? And the l2norm ,...

![image](https://user-images.githubusercontent.com/53176521/97283187-7f8dac80-187a-11eb-8c7e-d8d243e53c3c.png) ![1](https://user-images.githubusercontent.com/53176521/97575701-1a77b980-1a28-11eb-8713-76e7066d654a.png) ![2](https://user-images.githubusercontent.com/53176521/97577418-5f9ceb00-1a2a-11eb-82aa-c2ed432ef034.png) 2020.11.5 ![image](https://user-images.githubusercontent.com/53176521/98217023-d8e59200-1f84-11eb-982c-1d80425ad58d.png)

How to convert the label to voc form? Or,how should I use the label, is there any demo or code example?

请问一下大佬,如果训练自己的数据集,是不是dataloader返回的jpg和png就够用了,labels是针对voc数据集的白边的,再用diceloss的时候把参数里的label换成png?

![image](https://github.com/yangjian1218/seetaface6ToPy/assets/53176521/08c98314-b169-4f91-bc5a-4569d57ddf36) 这个应该如何解决呢