Question about KITTI raw dataset

[Len-Li]

Hi,
Thanks for sharing your work! I am wondering when will you release the dataset pipeline for KITTI raw and other datasets.
By the way, how to evaluate the network for each dataset? And what's the reported performance in LLFF dataset? I can't found them in the paper.
Thanks!

[vincentfung13]

Thank you for your interest in our work!

1. We will try to release the training flows for the other datasets by the end of the month.
2. We did not report the performance on the LLFF dataset as our setting is on single-view synthesis, which is quite different from the other baselines.

Hope this helps.

Zijian

[youngAt19]

@vincentfung13
Hi, in your paper you said that the model was trained on the 20 city sequences from the KITTI Raw dataset, and evaluated on another 4 city sequences.
Besides, MPI[51] and Tulsiani et. al. [52] reported that they used 22 city sequences for training and 4 for testing.
However, the dataset I downloaded from kitti contains more than 30 sequences.
Could you tell me which sequences do you used for training and testing?

[vincentfung13]

@youngAt19 Hi, following Tulsiani et. al. 2018 and Tucker et. al. 2020, we used the code here to generate our splits and report the results on the test split. Hope this helps.

[youngAt19]

@vincentfung13 Wow! Thanks a lot!!

[Len-Li]

Hi, could you release the remaining code for other datasets?

[vincentfung13]

@Len-Li releasing the data pipelines for the other datasets requires re-writing all the dataloaders since we used internal libraries from ByteDance so we haven't got time to prepare it. However please feel free to drop us an email to request code snippets should you have any questions regarding the preprocessing / data loaders for the datasets. My apologies for the inconvenience.

Zijian