search

szymanowiczs / splatter-image

Official implementation of `Splatter Image: Ultra-Fast Single-View 3D Reconstruction' CVPR 2024
https://szymanowiczs.github.io/splatter-image
BSD 3-Clause "New" or "Revised" License
863 stars 68 forks source link

Gaussian Renderer Issue #62

Open kunalkathare opened 2 weeks ago

kunalkathare commented 2 weeks ago

Hi, I'm getting the following error, when trying to run eval.py on shapenet cars dataset. using python 3.7, torch=1.12.1, cuda11.6 installed diff-gaussian-rasterizer from gaussian splatting official repo.

Evaluating on dataset cars Will load a model released with the paper. Not saving any renders (only computing scores). To save renders use flag --save_vis Loaded model! 704 Setting up [LPIPS] perceptual loss: trunk [vgg], v[0.1], spatial [off] /home/test/anaconda3/envs/splatter-image/lib/python3.7/site-packages/torchvision/models/_utils.py:209: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and will be removed in 0.15, please use 'weights' instead. f"The parameter '{pretrained_param}' is deprecated since 0.13 and will be removed in 0.15, " /home/test/anaconda3/envs/splatter-image/lib/python3.7/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or None for 'weights' are deprecated since 0.13 and will be removed in 0.15. The current behavior is equivalent to passing weights=VGG16_Weights.IMAGENET1K_V1. You can also use weights=VGG16_Weights.DEFAULT to get the most up-to-date weights. warnings.warn(msg) Loading model from: /home/test/anaconda3/envs/splatter-image/lib/python3.7/site-packages/lpips/weights/v0.1/vgg.pth 0%| | 0/704 [00:02<?, ?it/s] Traceback (most recent call last): File "eval.py", line 326, in scores = main(dataset_name, experiment_path, 0, split=split, save_vis=save_vis, out_folder=out_folder) File "/home/test/anaconda3/envs/splatter-image/lib/python3.7/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context return func(*args, *kwargs) File "eval.py", line 287, in main scores = evaluate_dataset(model, dataloader, device, training_cfg, save_vis=save_vis, out_folder=out_folder) File "/home/test/anaconda3/envs/splatter-image/lib/python3.7/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context return func(args, **kwargs) File "eval.py", line 112, in evaluate_dataset focals_pixels=focals_pixels_render)["render"] File "/data/kunal/splatter-image/gaussian_renderer/init.py", line 53, in render_predicted debug=False TypeError: new() missing 1 required positional argument: 'antialiasing'

Andrewwango commented 1 week ago

Annoyingly due to the unstable API of diff_gaussian_rasterizer, some changes need to be made in gaussian_renderer/__init__.py: (e.g. non-optional arg antialiasing introduced in this commit)

You will need to add antialiasing=True to https://github.com/szymanowiczs/splatter-image/blob/78a6ad098e0cdc40c59c8ec98ca4fa439870fabd/gaussian_renderer/__init__.py#L41

and change https://github.com/szymanowiczs/splatter-image/blob/78a6ad098e0cdc40c59c8ec98ca4fa439870fabd/gaussian_renderer/__init__.py#L84 to rendered_image, radii, _ = rasterizer(

Are you still actively developing this repo to keep up with the changes @szymanowiczs ? Thanks