yitu-opensource / T2T-ViT

ICCV2021, Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet
Other
1.14k stars 177 forks source link

the input channels #13

Closed DDxk369 closed 3 years ago

DDxk369 commented 3 years ago
    您好,非常感谢您开源这么棒的一个工作,我对t2t特别感兴趣,并且向将它应用于医疗方向的数据,但是这就需要把输入信道的维数从3更改为1,对您的工作正在学习当中,对很多代码还不是特别熟悉,所以在这里想问下您,在哪里可以设置输入信道的相关参数?
    谢谢!!!!
yuanli2333 commented 3 years ago

The default value of input channel for image is 3,you can change the 'in_chans=3' in this line.