Encountered an error during the first training session on D-Nerf dataset

[HsiehTPE]

Dear devs,

After I configured the training environment as required, I used

Train with terminal

for the first training(I have edited data path in the code): # Train with terminal only (for the resolution of 400*400 with best PSNR) CUDA_VISIBLE_DEVICES=0 python train_gui.py --source_path YOUR/PATH/TO/DATASET/jumpingjacks --model_path outputs/jumpingjacks --deform_type node --node_num 512 --hyper_dim 8 --is_blender --eval --gt_alpha_mask_as_scene_mask --local_frame --resolution 2 --W 800 --H 800

But encountered the following error type:

Below is the structure of the dataset I placed:

Could you please help me identify the possible reasons for this error? Thank you very much! Once again, thank you for your extraordinary work.

[yihua7]

Hi, It's because the Gaussian rasterization lib you use is not exactly what we recommend in the readme file.

You can choose to reinstall the rasterizer using: pip install ./submodules/diff-gaussian-rasterization or remove the alpha in the code since your current version of the rasterizer seems to return three values rendered_image, radii, depth only.

[HsiehTPE]

It works, thanks for your reply!

Wish you have a good day : )