How to generating multi-view images from other public datasets

[xwhkkk]

Hello!
thanks for sharing your great job!
I want to use the pretrain model to generate multi-view images with other public datasets. Could you give me some suggestions

[MartinMao1101]

hi there!
I am also trying to use this great model to process with my dataset in infrared sprectrum, could have a talk if you are also interested

[xwhkkk]

hi there! I am also trying to use this great model to process with my dataset in infrared sprectrum, could have a talk if you are also interested

Hello! What's your purpose to process the infrared spectrum dataset? I noticed that issue #73 may be helpful, but I have not tested it.

[MartinMao1101]

Hi thanks a lot for the best tip! It is quite helpful for my situation.
My situation is that, i have some infrared real human face imgs (in greyscale), and i want to generate some synthetic human face imgs (also in greyscale) with consistency from multiview by using EG3D. So i guess we are in similar situation, difference is that i need infrared imgs instead of RGB imgs.
Btw i also saw the #29 maybe could answer your question here

[xwhkkk]

Hi thanks a lot for the best tip! It is quite helpful for my situation. My situation is that, i have some infrared real human face imgs (in greyscale), and i want to generate some synthetic human face imgs (also in greyscale) with consistency from multiview by using EG3D. So i guess we are in similar situation, difference is that i need infrared imgs instead of RGB imgs. Btw i also saw the #29 maybe could answer your question here

Hi thanks a lot for the best tip! It is quite helpful for my situation. My situation is that, i have some infrared real human face imgs (in greyscale), and i want to generate some synthetic human face imgs (also in greyscale) with consistency from multiview by using EG3D. So i guess we are in similar situation, difference is that i need infrared imgs instead of RGB imgs. Btw i also saw the #29 maybe could answer your question here

Thank you! Have you figured out how to pass the input images ?

[MartinMao1101]

According to the comment, my understanding would be, that this is an unconditional model, so when using it, you cannot pass an extra input img into the model, the generation is based on some other params like seeds.
But of course, when doing the training, you need to feed the model with some imgs by defining the hyperparam --data=path_to_your_training_dataset

[povolann]

Hi thanks a lot for the best tip! It is quite helpful for my situation. My situation is that, i have some infrared real human face imgs (in greyscale), and i want to generate some synthetic human face imgs (also in greyscale) with consistency from multiview by using EG3D. So i guess we are in similar situation, difference is that i need infrared imgs instead of RGB imgs. Btw i also saw the #29 maybe could answer your question here

Hi, I see you are using also grayscale images, could you possibly look at my comment here? Thank you!

[yanre-hyd]

在进行训练时，您需要通过定义 hyperparam --data=path_to_your_training_dataset 为模型提供一些 imgs

你好，请问你解决了这个不能传入img到模型中的这个问题吗，我需要将img传入模型生成自己的图像