How to evaluate the codes returning from the autoencoder? about meshgpt-pytorch HOT 6 CLOSED

@adeerAI give me until end of the month. the repository will be in a good place by then

from meshgpt-pytorch.

@fire looks good! A for effort 🥇

made it a bit easier with a keyword argument on forward

hope we start seeing some results soon!

from meshgpt-pytorch.

@fire @lucidrains Thank you both for your wonderful work and dedication. I just wanna ask, as to when it can be ready to train and test it for models. Also, can anyone of you make like a proper User guide file on how to train with a dataset and how to use it for mesh making using text. Thanks a lot.

from meshgpt-pytorch.

@fire if you set this keyword arg to True, it will return the reconstructed face coordinates

from meshgpt-pytorch.

Can you try again? The git link moved.

from meshgpt-pytorch.

transformer_trainer.train(run.config.transformer_train)

codes, continuous_coors = transformer.generate(return_codes=True)

codes_list = codes.cpu().tolist()

import json

with open("output_codes.json", "w") as f:
    json.dump(codes_list, f)

continuous_coors_list = continuous_coors.cpu().tolist()

with open("continuous_coors.json", "w") as f:
    json.dump(continuous_coors.tolist(), f)

flat_list = [item for sublist in continuous_coors_list for item in sublist]

vertices = [vertex for sublist in flat_list for vertex in sublist]

faces = [[i, i + 1, i + 2] for i in range(0, len(vertices), 3)]

Here's my tweaked code. Thanks

from meshgpt-pytorch.