Hello, when I try to run "Spatial Temporal Interpolation Visualizations" on the benchmark of nvidia dynamic scenes, I got a problem like this:

gnt: 91%|█████████ | 70/77 [2:55:22<17:32, 150.33s/it] gnt: 92%|█████████▏| 71/77 [2:57:54<15:04, 150.74s/it] gnt: 94%|█████████▎| 72/77 [3:00:22<12:30, 150.04s/it] gnt: 95%|█████████▍| 73/77 [3:02:54<10:02, 150.56s/it] gnt: 96%|█████████▌| 74/77 [3:05:23<07:30, 150.10s/it] gnt: 97%|█████████▋| 75/77 [3:07:51<04:58, 149.32s/it] gnt: 99%|█████████▊| 76/77 [3:10:19<02:29, 149.11s/it] gnt: 100%|██████████| 77/77 [3:12:23<00:00, 149.92s/it] Error executing job with overrides: ['verbose=true', 'distributed=false', 'seed=0', 'resume=vis_wo_resume', 'resume_dir=null', 'engine=visualizer_pgdvs', 'model=pgdvs_renderer', 'model.softsplat_metric_abs_alpha=100.0', 'static_renderer=gnt', 'static_renderer.model_cfg.ckpt_path=/data/code/ml-pgdvs/ckpts/gnt/model_720000.pth', 'series_eval=false', 'eval_batch_size=1', 'n_max_eval_data=-1', 'eval_save_individual=true', 'engine.engine_cfg.render_cfg.render_stride=1', 'engine.engine_cfg.render_cfg.chunk_size=2048', 'engine.engine_cfg.render_cfg.sample_inv_uniform=true', 'engine.engine_cfg.render_cfg.n_coarse_samples_per_ray=256', 'engine.engine_cfg.render_cfg.n_fine_samples_per_ray=0', 'engine.engine_cfg.render_cfg.mask_oob_n_proj_thres=1', 'engine.engine_cfg.render_cfg.mask_invalid_n_proj_thres=4', 'engine.engine_cfg.render_cfg.dyn_pcl_remove_outlier=true', 'engine.engine_cfg.render_cfg.dyn_pcl_outlier_knn=50', 'engine.engine_cfg.render_cfg.dyn_pcl_outlier_std_thres=0.1', 'engine.engine_cfg.render_cfg.gnt_use_dyn_mask=true', 'engine.engine_cfg.render_cfg.gnt_use_masked_spatial_src=false', 'engine.engine_cfg.render_cfg.dyn_render_use_flow_consistency=false', 'dataset=combined', 'dataset.dataset_list.train=[nvidia_eval]', 'dataset.dataset_list.eval=[nvidia_eval]', 'dataset.dataset_list.vis=[nvidia_vis]', 'dataset.dataset_specifics.mono_vis.scene_ids=[Balloon1]', 'dataset.data_root=/data/code/ml-pgdvs/data', 'n_dataloader_workers=1', 'dataset_max_hw=-1', 'dataset.use_aug=false', 'dataset.dataset_list.vis=[nvidia_vis]', 'dataset.dataset_specifics.nvidia_vis.scene_ids=[Balloon1]', 'vis_specifics.n_render_frames=400', 'vis_specifics.vis_center_time=50', 'vis_specifics.vis_time_interval=50', 'vis_specifics.vis_bt_max_disp=32'] Traceback (most recent call last): File "/data/code/ml-pgdvs/pgdvs/run.py", line 267, in <module> cli() File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/main.py", line 90, in decorated_main _run_hydra( File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/_internal/utils.py", line 389, in _run_hydra _run_app( File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/_internal/utils.py", line 452, in _run_app run_and_report( File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/_internal/utils.py", line 216, in run_and_report raise ex File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/_internal/utils.py", line 213, in run_and_report return func() File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/_internal/utils.py", line 453, in <lambda> lambda: hydra.run( File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/_internal/hydra.py", line 132, in run _ = ret.return_value File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/core/utils.py", line 260, in return_value raise self._return_value File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/hydra/core/utils.py", line 186, in run_job ret.return_value = task_function(task_cfg) File "/data/code/ml-pgdvs/pgdvs/run.py", line 263, in cli run(cfg, hydra_config) File "/data/code/ml-pgdvs/pgdvs/run.py", line 192, in run return _distributed_worker( File "/data/code/ml-pgdvs/pgdvs/run.py", line 146, in _distributed_worker output = engine.run() File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context return func(*args, **kwargs) File "/data/code/ml-pgdvs/pgdvs/engines/visualizer_pgdvs.py", line 26, in run self.vis_model() File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context return func(*args, **kwargs) File "/data/code/ml-pgdvs/pgdvs/engines/visualizer_pgdvs.py", line 91, in vis_model ret_dict = self._get_model_module(self.model).forward( File "/data/code/ml-pgdvs/pgdvs/renderers/pgdvs_renderer.py", line 146, in forward (render_dyn_rgb, render_dyn_mask, render_dyn_info) = self.dyn_renderer( File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1518, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1527, in _call_impl return forward_call(*args, **kwargs) File "/data/code/ml-pgdvs/pgdvs/renderers/pgdvs_renderer_dyn.py", line 184, in forward splat_dyn_img_full, softsplat_metric_src1_to_src2 = self.softsplat_img( File "/data/code/ml-pgdvs/pgdvs/renderers/pgdvs_renderer_base.py", line 80, in softsplat_img splat_img_src1_to_tgt = softsplat.softsplat( File "/data/code/ml-pgdvs/pgdvs/utils/softsplat.py", line 311, in softsplat tenOut = softsplat_func.apply(tenIn, tenFlow) File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/torch/autograd/function.py", line 539, in apply return super().apply(*args, **kwargs) # type: ignore[misc] File "/data/apps/anaconda3/envs/pgdvs/lib/python3.10/site-packages/torch/cuda/amp/autocast_mode.py", line 121, in decorate_fwd return fwd(*args, **kwargs) File "/data/code/ml-pgdvs/pgdvs/utils/softsplat.py", line 421, in forward assert False AssertionError

I have followed the instruction of "https://github.com/apple/ml-pgdvs/blob/main/docs/BENCHMARK_NVIDIA.md" , and I just want to show the visuable result. I guess this may be related to the config process, but how could I solve this problem? Thank you!

Environment: CUDA 11.8 + nvidia_A800 and without jax

Hello, I notice that there is a pretrained gnt model in your download_ckpts.sh like this:

if [ "${FLAG_ORIGINAL}" == "1" ]; then
# GNT
if [ ! -f ${DATA_ROOT}/gnt/generalized_model_720000.pth ]; then
gdown 1AMN0diPeHvf2fw53IO5EE2Qp4os5SkoX -O ${DATA_ROOT}/gnt/
fi

However, I couldn't find this file on google drive and they tell me it not exist. How could I get this?
Thanks!

Since the point clouds are filtered by a cycle consistent check, how do you inpaint the filtered points?

I notice that you project the point cloud to the image plane for rendering. But how can you make sure each pixel has its corresponding point cloud? Are there any constrains?

Hello,

I was trying to load the DyCheck dataset in another framework and stumbled upon your fixes to a DyCheckCamera class.

I seem to be understanding that you assume OpenCV conventions everywhere after you, for example, call cam.extrin and obtain the extrinsics matrix (according to comments world-to-camera).

However, when you call the extrin property note that you also return the translation as -orientation @ position, implying that the position is in camera-to-world and you're taking the inverse of the stored position as $-R^\top t$ -- and the orientation is coherently assumed to be stored as camera-to-world, so that $R^\top$ is its inverse.

Have you confirmed this system is consistent with the expected trajectory, e.g., on the paper-windmill example from the iphone dataset in dycheck-release?