diamond

What are the attributes of segmap for every class? Thank you！

when traing, the loss becomes nan. why?

2


You give 9 scirpts for training. Can you show the different performance of the 9 scripts for us? Thank you!

I want to see the accuracy of the warping. So how can i see the warped image during the training stage. I want to make sure whether the source image...

About the prompt rules

11

Great work! I would like to know what kind of sentences are reasonable and valid for CLIP. Are there any specific prompt rules for style sentences in your paper? I...

Hulk performance in Figure 3.

1

I train the code and use the default script 'python main.py --run branch --obj_path data/source_meshes/person.obj --output_dir results/demo/people/hulk --prompt "a 3D rendering of the Hulk in unreal engine" --sigma 12.0 --clamp...

Should the four places below be normloss? I don't think the loss is passed back? https://github.com/threedle/text2mesh/blob/07570f7858c341d72e316f084c7a1973b205cc62/main.py#L250 https://github.com/threedle/text2mesh/blob/07570f7858c341d72e316f084c7a1973b205cc62/main.py#L253 https://github.com/threedle/text2mesh/blob/07570f7858c341d72e316f084c7a1973b205cc62/main.py#L293 https://github.com/threedle/text2mesh/blob/07570f7858c341d72e316f084c7a1973b205cc62/main.py#L296

why do you split the celeba-HQ to train and val set? It seems like the val set is not used.

I would like to ask how to determine what semantic information each sub-generator is responsible for generating. There doesn't seem to be any oversight or enforcement of which module each...