yukangcao / sesdf Goto Github PK

Hi,
I know that SeSDF can train shape, but I did not find a way to train color.

Does SeSDF support color/texture training from images?

Regards

Hi, thanks for your great work.
I m trying to train with my own data.
I finished dataset preprocessing. but when I try to train the shape, the program hands up.
Even wait for a day, it still hands up there.

my command is:
python -m apps.train_shape --dataroot /mnt/d/wang/windows/repos/SeSDF/custom_data/generated --random_flip --random_scale --resolution 512 --hg_down 'ave_pool' --norm 'group' --val_train_error --val_test_error --gpu_ids=0 --batch_size 2 --learning_rate 0.0001 --sigma 0.02 --checkpoints_path /mnt/d/wang/windows/repos/SeSDF/custom_data/checkpoints --results_path /mnt/d/wang/windows/repos/SeSDF/custom_data/result

the output is as the following:

train data size: 180
test data size: 360
initialize network with normal
Using Network: hgpifu

it looks like it hangs up at this line of the code

my env is as the following

      - torch==2.1.1+cu121
      - torch-scatter==2.1.2+pt21cu121
      - torchmetrics==0.9.1
      - torchvision==0.16.1+cu121
      - pytorch-lightning==1.6.4
      - pytorch3d==0.7.6
      - kaolin==0.15.0
      - python=3.8.13=h12debd9_0

my gpu is GX 4090.

May I ask if you can provide the trained model weights。thank you

Good job! Could you share the inference code of multi-view pipeline?