Re10k_subset numerical evaluation

[Gaozihui]

Hi, congratulations on your work!

I got results of re10k_subset in outputs/test and when I run python3 -m src.scripts.compute_metrics +experiment=re10k +evaluation=re10k dataset.roots=[datasets/re10k_subset] it outputs skipping and without a numerical eval: Testing: | | 0/? [00:00<?, ?it/s]Skipping "5aca87f95a9412c6". Skipping "322261824c4a3003". Testing DataLoader 0: 0%| | 0/41 [00:00<?, ?it/s]Skipping "5aca87f95a9412c6". Testing DataLoader 0: 2%|██▍ | 1/41 [00:00<00:00, 1084.64it/s]Skipping "322261824c4a3003". Testing DataLoader 0: 5%|████▉ | 2/41 [00:00<00:00, 475.19it/s]Skipping "17d9303ee77c3a3d". Testing DataLoader 0: 7%|███████▍ | 3/41 [00:00<00:00, 377.63it/s]Skipping "17d9303ee77c3a3d". Skipping "bc95e5c7e357f1b7". Skipping "f7c0fa5b81552d35". Skipping "ffa95c3b40609c76". Skipping "bc95e5c7e357f1b7". Testing DataLoader 0: 10%|█████████▉ | 4/41 [00:00<00:00, 50.82it/s]Skipping "f7c0fa5b81552d35". Testing DataLoader 0: 12%|████████████▍ | 5/41 [00:00<00:00, 60.90it/s]Skipping "ffa95c3b40609c76". Testing DataLoader 0: 15%|██████████████▉ | 6/41 [00:00<00:00, 70.36it/s]

[Gaozihui]

Also, I run python3 -m src.scripts.generate_evaluation_index +experiment=re10k +evaluation=re10k dataset.roots=[datasets/re10k_subset] to generate re10k_subset index but I have no idea how to use it.

[dcharatan]

For evaluation, you'll need the full dataset (see the note in the README—email us to obtain it).

You would only need to generate a new evaluation index if you're running pixelSplat on your own dataset. In that case, it's used with this configuration item. You can change the view sampler/this configuration from the command line. For example, here's the ACID setup from the README:

python3 -m src.main +experiment=acid mode=test dataset/view_sampler=evaluation dataset.view_sampler.index_path=assets/evaluation_index_acid.json checkpointing.load=checkpoints/acid.ckpt

[CBoLee]

For evaluation, you'll need the full dataset (see the note in the README—email us to obtain it).

You would only need to generate a new evaluation index if you're running pixelSplat on your own dataset. In that case, it's used with this configuration item. You can change the view sampler/this configuration from the command line. For example, here's the ACID setup from the README:

python3 -m src.main +experiment=acid mode=test dataset/view_sampler=evaluation dataset.view_sampler.index_path=assets/evaluation_index_acid.json checkpointing.load=checkpoints/acid.ckpt

hello

For evaluation, you'll need the full dataset (see the note in the README—email us to obtain it).

You would only need to generate a new evaluation index if you're running pixelSplat on your own dataset. In that case, it's used with this configuration item. You can change the view sampler/this configuration from the command line. For example, here's the ACID setup from the README:

python3 -m src.main +experiment=acid mode=test dataset/view_sampler=evaluation dataset.view_sampler.index_path=assets/evaluation_index_acid.json checkpointing.load=checkpoints/acid.ckpt

hello，thank your excellent work，I'm interested in accessing the complete RealEstate10K dataset for my research. Could you please provide access details?my Email is licypress0102@gmail.com