[Dataset] How to create the dataset with my own photos？

[xdobetter]

Hi, very interesting project. I ran your project, and it was great！ But I have some questions: how do I make my own dataset to fit your project?

[YuliangXiu]

For testing, you can use any photos with the same clothing, please take a look at the photos I used to create the teaser figure via

wget https://download.is.tue.mpg.de/icon/human.zip

To benchmark, you need to prepare the photos, together with the A-posed textured scans (incl. A-posed SMPL-X)

[xdobetter]

I appreciate your help and I have successfully run it on my own data. I have an additional question. The content in my current gpt4v_simple.json is obtained by querying ChatGPT-4o using the prompt provided in the paper. Do you have a corresponding script in your project to obtain this content?

[YuliangXiu]

Is grounding_dino_sam.py#L219 what you asked for? The gpt4v_simple.json should be there if you run grounding_dino_sam.py as in run.sh#L12

[YuliangXiu]

Okay, I noticed that I couldn't upload the gpt4v_complex.txt and gpt4v_simple.txt files because of .gitignore. I'll update them tonight and let you know.

[xdobetter]

Is grounding_dino_sam.py#L219 what you asked for? The gpt4v_simple.json should be there if you run grounding_dino_sam.py as in run.sh#L12

Yes, that's exactly what I want to ask. I use the gpt4v_prompt.txt you provided instead of gpt4v_simple.txt, which can automatically generate gpt4v_simple.json. I don't know if your code is a bug in the line https://github.com/YuliangXiu/PuzzleAvatar/blob/e1d46171f3f8cdada56b22d556f2cdfc291f446f/multi_concepts/grounding_dino_sam.py#L219

[YuliangXiu]

I have updated both gpt4v_simple.txt and gpt4v_complex.txt.