Hang Su

Any plan on when these will be implemented? Thanks for sharing the awesome work!

Check how we handle data loading in splatnet/splatnet/dataset/dataset_facade.py. Just don't include "h" dimension if you don't want to use height. Also if you use custom data files, you will need...

Thanks for your contribution. I've added the information in the readme.

This [line of code](https://github.com/NVlabs/splatnet/blob/master/splatnet/plot_log.py#L20) calls the caffe script (`tools/extra/parse_log.py`) to parse the .log file into .log.train and .log.test. The number of samples shouldn't matter here. You can try to run...

I am facing the same issue. The number I got is 0.04647076057203757. Using the provided .pth file without any code modification.

Thanks, @h982639009 ! After that change, I got: `pytorch 20 Mean iou = 0.7645974317511359`. This is even higher than the number 76.35 reported in the [paper](https://arxiv.org/pdf/1606.00915.pdf) (Table. 4 row 6)....

I did the fix @h982639009 mentioned, and made no other changes.

Were the data files generated as instructed [here](https://github.com/NVlabs/splatnet/tree/master/data#ruemonge428) using the same machine? There is no specified codec when generating the data files, but I would assume loading the files written...

I am extremely sorry for the inconvenience ... This is still in the plan, but we won't be able to provide a reliable delivery time as there are a few...

To add a dataset properly, you will need to modify files under `splatnet/semseg3d/` (or `splatnet/partseg3d/` for part segmentation) accordingly. A quicker solution is to keep the original code but instead...