xiyufeng2

How to use it with text prompt? I noticed that there is only point and box prompt in the code.

checkpoint mismatch

1

when I load the trained checkpoint， there is an error： size mismatch for prop_mlp_0.encoder.embeddings: copying a param with shape torch.Size([6606952, 1]) from checkpoint, the shape in current model is torch.Size([6606952,...

low quality

4

there is always too much floaters in rendered img. what should I do when I train my data? Thank you!

Segmentation fault (core dumped)

1

python train.py It worked on ubuntu with cuda11.7,but failed with another computer( cuda11.3) the errer is "Segmentation fault (core dumped)"

the input data(32,3,1024) is input into the network, and the result's demension is (32,40).1024 points is missing.what a bad code.

Hello,author. There is 4 pretrained model provided, including chair, car ,etc. How to train my own model, or just follow the GET3D?

Run end-to-end optimazation, Found no checkpoint file in ...experiments/real/001 with prefix checkpoint. checkpoint step 0

mesh没有纹理信息，只有几何信息

1

将GS提取成mesh，结果没有纹理信息，是黑白的mesh。我的输入就是原始GS的结果，为啥啊