Zhang Yu

the question about the loss function

5

Hi， @meteorshowers ！ Thanks for your work, it's useful for me. When I read the code, I find that in `train_ent_disp.py` you import `My_loss` from the `disparity.models.loss` module, but I...

Hi, thank for you work! It's very important for me. If I want to use my own pictures to inference, is there a requirement for the size of the pictures?

The question about train.py

2

When I run the `train.py` file, the results of `validate_things` in it is Nan, but when I run `validate_things` alone, it can output epe and d1. So why?

Hello, thanks for your code, it is useful for me! Can you please upload the usage of this repository? It will be convenient for me. Thanks!

关于网络模型泛化能力的问题

4

您好，感谢您开源的代码 对于网络结构，我想问一下，您是如何提高模型的泛化能力的，我使用了其他网络模型来预测我自己拍的图像，但是效果十分差，但是用了您的IGEV网络跑出来的效果还是比较可观的，想咨询一下您是如何做到这么强的泛化性的。谢谢

Great Work! When will you release the code? Thanks